Mr. Justice Powell

delivered the opinion of the Court.

In this case, brought under 42 U. S. C. § 1983, we consider the elements and prerequisites for recovery of damages by students who were suspended from public elementary and secondary schools without procedural due process. The Court of Appeals for the Seventh Circuit held that the students are entitled to recover substantial nonpunitive damages even if their suspensions were justified, and even if they do not prove that any other actual injury was caused by the denial of procedural due process. We disagree, and hold that in the absence of proof of actual injury, the students are entitled to recover only nominal damages.

I

Respondent Jarius Piphus was a freshman at Chicago Vocational High School during the 1973-1974 school year. On January 23, 1974, during school hours, the school principal saw Piphus and another student standing outdoors on school property passing back and forth what the principal described as an irregularly shaped cigarette. The principal approached the students unnoticed and smelled what he believed was the *249 strong odor of burning marihuana. He also saw Piphus try to pass a packet of cigarette papers to the other student. When the students became aware of the principal’s presence, they threw the cigarette into a nearby hedge.

The principal took the students to the school’s disciplinary office and directed the assistant principal to impose the “usual” 20-day suspension for violation of the school rule against the use of drugs. 1 The students protested that they had not been smoking marihuana, but to no avail. Piphus was allowed to remain at school, although not in class, for the remainder of the school day while the assistant principal tried, without success, to reach his mother.

A suspension notice was sent to Piphus’ mother, and a few days later two meetings were arranged among Piphus, his mother, his sister, school officials, and representatives from a legal aid clinic. The purpose of the meetings was not to determine whether Piphus had been smoking marihuana, but rather to explain the reasons for the suspension. Following an unfruitful exchange of views, Piphus and his mother, as guardian ad litem, filed suit against petitioners in Federal District Court under 42 U. S. C. § 1983 and its jurisdictional *250 counterpart, 28 U. S. C. § 1343, charging that Piphus had been suspended without due process of law in violation of the Fourteenth Amendment. The complaint sought declaratory and injunctive relief, together with actual and punitive damages in the amount of $3,000. 2 Piphus was readmitted to school under a temporary restraining order after eight days of his suspension.

Respondent Silas Brisco was in the sixth grade at Clara Barton Elementary School in Chicago during the 1973-1974 school year. On September 11, 1973, Brisco came to school wearing one small earring. The previous school year the school principal had issued a rule against the wearing of earrings by male students because he believed that this practice denoted membership in certain street gangs and increased the likelihood that gang members would terrorize other students. Brisco was reminded of this rule, but he refused to remove the earring, asserting that it was a symbol of black pride, not of gang membership.

The assistant principal talked to Brisco’s mother, advising her that her son would be suspended for 20 days if he did not remove the earring. Brisco’s mother supported her son’s position, and a 20-day suspension was imposed. Brisco and his mother, as guardian ad litem, filed suit in Federal District Court under 42 U. S. C. § 1983 and 28 U. S. C. § 1343, charging that Brisco had been suspended without due process of law in violation of the Fourteenth Amendment. 3 The complaint *251 sought declaratory and injunctive relief, together with actual and punitive damages in the amount of $5,000. 4 Brisco was readmitted to school during the pendency of proceedings for a preliminary injunction after 17 days of his suspension.

Piphus’ and Brisco’s cases were consolidated for trial and submitted on stipulated records. The District Court held that both students had been suspended without procedural due process. 5 It also held that petitioners were not entitled to qualified immunity from damages under the second branch of Wood v. Strickland, 420 U. S. 308 (1975), because they “should have known that a lengthy suspension without any adjudicative hearing of any type” would violate procedural due process. App. to Pet. for Cert. A14. 6 Despite these holdings, the District Court declined to award damages because:

“Plaintiffs put no evidence in the record to quantify their *252 damages, and the record is completely devoid of any evidence which could even form the basis of a speculative inference measuring the extent of their injuries. Plaintiffs’ claims for damages therefore fail for complete lack of proof.” Ibid.

The court also stated that the students were entitled to declaratory relief and to deletion of the suspensions from their school records, but for reasons that are not apparent the court failed to enter an order to that effect. Instead, it simply dismissed the complaints. No finding was made as to whether respondents would have been suspended if they had received procedural due process.

On respondents’ appeal, the Court of Appeals reversed and remanded. 545 F. 2d 30 (1976). It first held that the District Court erred in not granting declaratory and injunctive relief. It also held that the District Court should have considered evidence submitted by respondents after judgment that tended to prove the pecuniary value of each day of school that they missed while suspended. The court said, however, that respondents would not be entitled to recover damages representing the value of missed school time if petitioners showed on remand “that there was just cause for the suspension [s] and that therefore [respondents] would have been suspended even if a proper hearing had been held.” Id., at 32.

Finally, the Court of Appeals held that even if the District Court found on remand that respondents’ suspensions were justified, they would be entitled to recover substantial “non-punitive” damages simply because they had been denied procedural due process. Id., at 31. Relying on its earlier *253 decision in Hostrop v. Board of Junior College Dist. No. 515, 523 F. 2d 569 (CA7 1975), cert. denied, 425 U. S. 963 (1976), the court stated that such damages should be awarded “even if, as in the case at bar, there is no proof of individualized injury to the plaintiff, such as mental distress . . . .” 545 F: 2d, at 31. We granted certiorari to consider whether, in an action under § 1983 for the deprivation of procedural due process, a plaintiff must prove that he actually was injured by the deprivation before he may recover substantial “non-punitive” damages. 430 U. S. 964 (1977).

II

Title 42 U. S. C. § 1983, Rev. Stat. § 1979, derived from § 1 of the Civil Rights Act of 1871, 17 Stat. 13, provides:

“Every person who, under color of any statute, ordinance, regulation, custom, or usage, of any State or Territory, subjects, or causes to be subjected, any citizen of the United States or other person within the jurisdiction thereof to the deprivation of any rights, privileges, or immunities secured by the Constitution and laws, shall be liable to the party injured in an action at law, suit in equity, or other proper proceeding for redress.”

The legislative history of § 1983, elsewhere detailed, e. g., Monroe v. Pape, 365 U. S. 167, 172-183 (1961); id., at 225-234 (Frankfurter, J., dissenting in part); Mitchum v. Foster, 407 U. S. 225, 238-242 (1972), demonstrates that it was intended to “[create] a species of tort liability” in favor of persons who are deprived of “rights, privileges, or immunities secured” to them by the Constitution. Imbler v. Pachtman, 424 U. S. 409, 417 (1976).

Petitioners contend that the elements and prerequisites for recovery of damages under this “species of tort liability” should parallel those for recovery of damages under the common law of torts. In particular, they urge that the purpose of an award of damages under § 1983 should be to compensate *254 persons for injuries that are caused by the deprivation of constitutional rights; and, further, that plaintiffs should be required to prove not only that their rights were violated, but also that injury was caused by the violation, in order to recover substantial damages. Unless respondents prove that they actually were injured by the deprivation of procedural due process, petitioners argue, they are entitled at most to nominal damages.

Respondents seem to make two different arguments in support of the holding below. First, they contend that substantial damages should be awarded under § 1983 for the deprivation of a constitutional right whether or not any injury was caused by the deprivation. This, they say, is appropriate both because constitutional rights are valuable in and of themselves, and because of the need to deter violations of constitutional rights. Respondents believe that this view reflects accurately that of the Congress that enacted § 1983. Second, respondents argue that even if the purpose of a § 1983 damages award is, as petitioners contend, primarily to compensate persons for injuries that are caused by the deprivation of constitutional rights, every deprivation of procedural due process may be presumed to cause some injury. This presumption, they say, should relieve them from the necessity of proving that injury actually was caused.

A

Insofar as petitioners contend that the basic purpose of a § 1983 damages award should be to compensate persons for injuries caused by the deprivation of constitutional rights, they have the better of the argument. Rights, constitutional and otherwise, do not exist in a vacuum. Their purpose is to protect persons from injuries to particular interests, and their contours are shaped by the interests they protect.

Our legal system’s concept of damages reflects this view of legal rights. “The cardinal principle of damages in Anglo- *255 American law is that of compensation for the injury caused to plaintiff by defendant’s breach of duty.” 2 F. Harper & F. James, Law of Torts §25.1, p. 1299 (1956) (emphasis in original). 7 The Court implicitly has recognized the applicability of this principle to actions under § 1983 by stating that damages are available under that section for actions “found ... to have been violative of . . . constitutional rights and to have caused compensable-injury . . . .” Wood v. Strickland, 420 U. S., at 319 (emphasis supplied); see C-odd v. Velger, 429 U. S. 624, 630-631 (1977) (Brennan, J., dissenting); Adickes v. S. H. Kress & Co., 398 U. S. 144, 232 (1970) (Brennan, J., concurring and dissenting); see also Bivens v. Six Unknown Fed. Narcotics Agents, 403 U. S. 388, 397 (1971) (action for damages directly under Fourth Amendment); id., at 408-409 (Harlan, J., concurring in judgment). The lower federal courts appear generally to agree that damages awards under § 1983 should be determined by the compensation principle. 8

The Members of the Congress that enacted § 1983 did not address directly the question of damages, but the principle that damages are designed to compensate persons for injuries caused by the deprivation of rights hardly could have been foreign to the many lawyers in Congress in 1871. 9 Two other *256 sections of the Civil Rights Act of 1871 appear to incorporate this principle, and no reason suggests itself for reading § 1983 differently. 10 To the extent that Congress intended that awards under § 1983 should deter the deprivation of constitutional rights, there is no evidence that it meant to establish a deterrent more formidable than that inherent in the award of *257 compensatory damages. See Imbler v. Pachtman, 424 U. S., at 442 (White, J., concurring in judgment). 11

B

It is less difficult to conclude that damages awards under § 1983 should be governed by the principle of compensation than it is to apply this principle to concrete cases. 12 But over the centuries the common law of torts has developed a set of rules to implement the principle that a person should be compensated fairly for injuries caused by the violation of his legal rights. These rules, defining the elements of damages *258 and the prerequisites for their recovery, provide the appropriate starting point for the inquiry under § 1983 as well. 13

It is not clear, however, that common-law tort rules of damages will provide a complete solution to the damages issue in every § 1983 case. In some cases, the interests protected by a particular branch of the common law of torts may parallel closely the interests protected by a particular constitutional right. In such cases, it may be appropriate to apply the tort rules of damages directly to the § 1983 action. See Adickes v. S. H. Kress & Co., 398 U. S., at 231-232 (Brennan, J., concurring and dissenting). In other cases, the interests protected by a particular constitutional right may not also be protected by an analogous branch of the common law of torts. See Monroe v. Pape, 365 U. S., at 196, and n. 5 (Harlan, J., concurring ) id., at 250-251 (Frankfurter, J., dissenting in part); Adickes v. S. H. Kress & Co., supra, at 232 (Brennan, J., concurring and dissenting); Bivens v. Six Unknown Fed. Narcotic Agents, 403 U. S., at 394; id., at 408-409 (Harlan, J., concurring in judgment). In those cases, the task will be the more difficult one of adapting common-law rules of damages to provide fair compensation for injuries caused by the deprivation of a constitutional right.

Although this task of adaptation will be one of some delicacy — as this case demonstrates — it must be undertaken. The purpose of § 1983 would be defeated if injuries caused by the deprivation of constitutional rights went uncompensated simply because the common law does not recognize an analogous cause of action. Cf. Jones v. Hildebrant, 432 U. S. 183, 190-191 (1977) (White, J., dissenting); Sullivan v. Little Hunting Park, 396 U. S. 229, 240 (1969). In order to further *259 the purpose of § 1983, the rules governing compensation for injuries caused by the deprivation of constitutional rights should be tailored to the interests protected by the particular right in question — just as the common-law rules of damages themselves were defined by the interests protected in the various branches of tort law. We agree with Mr. Justice Harlan that “the experience of judges in dealing with private [tort] claims supports the conclusion that courts of law are capable of making the types of judgment concerning causation and magnitude of injury necessary to accord meaningful compensation for invasion of [constitutional] rights.” Bivens v. Six Unknown Fed. Narcotics Agents, supra, at 409 (Harlan, J., concurring in judgment). With these principles in mind, we now turn to the problem of compensation in the case at hand.

C

The Due Process Clause of the Fourteenth Amendment provides:

“[N]or shall any State deprive any person of life, liberty, or property, without due process of law . . . .”

This Clause “raises no impenetrable barrier to the taking of a person’s possessions,” or liberty, or life. Fuentes v. Shevin, 407 U. S. 67, 81 (1972). Procedural due process rules are meant to protect persons not from the deprivation, but from the mistaken or unjustified deprivation of life, liberty, or property. Thus, in deciding what process constitutionally is due in various contexts, the Court repeatedly has emphasized that “procedural due process rules are shaped by the risk of error inherent in the truth-finding process . . . .” Mathews v. Eldridge, 424 U. S. 319, 344 (1976). 14 Such rules “mini *260 mize substantively unfair or mistaken deprivations of” life, liberty, or property by enabling persons to contest the basis upon which a State proposes to deprive them of protected interests. Fuentes v. Shevin, supra, at 81.

In this case, the Court of Appeals held that if petitioners can prove on remand that “[respondents] would have been suspended even if a proper hearing had been held,” 545 F. 2d, at 32, then respondents will not be entitled to recover damages to compensate them for injuries caused by the suspensions. The court thought that in such a case, the failure to accord procedural due process could not properly be viewed as the cause of the suspensions. Ibid.; cf. Mt. Healthy City Board of Ed. v. Doyle, 429 U. S. 274, 285-287 (1977); Arlington Heights v. Metropolitan Housing Dev. Corp., 429 U. S. 252, 270-271, n. 21 (1977). The court suggested that in such circumstances, an award of damages for injuries caused by the suspensions would constitute a windfall, rather than compensation, to respondents. 545 F. 2d, at 32, citing Hostrop v. Board of Junior College Dist. No. 515, 523 F. 2d, at 579; cf. Mt. Healthy City Board of Ed. v. Doyle, supra, at 285-286. We do not understand the parties to disagree with this conclusion. Nor do we. 15

The parties do disagree as to the further holding of the Court of Appeals that respondents are entitled to recover substantial — although unspecified — damages to compensate them for “the injury which is 'inherent in the nature of the *261 wrong'” 545 F. 2d, at 31, even if their suspensions were justified and even if they fail to prove that the denial of procedural due process actually caused them some real, if intangible, injury. Respondents, elaborating on this theme, submit that the holding is correct because injury fairly may be “presumed” to flow from every denial of procedural due process. Their argument is that in addition to protecting against unjustified deprivations, the Due Process Clause also guarantees the “feeling of just treatment” by the government. Anti-Fascist Committee v. McGrath, 341 U. S. 123, 162 (1951) (Frankfurter, J., concurring). They contend that the deprivation of protected interests without procedural due process, even where the premise for the deprivation is not erroneous, inevitably arouses strong feelings of mental and emotional distress in the individual who is denied this “feeling of just treatment.” They analogize their case to that of defamation per se, in which “the plaintiff is relieved from the necessity of producing any proof whatsoever that he has been injured” in order to recover substantial compensatory damages. C. McCormick, Law of Damages § 116, p. 423 (1935). 16

*262 Petitioners do not deny that a purpose of procedural due process is to convey to the individual a feeling that the government has dealt with him fairly, as well as to minimize the risk of mistaken deprivations of protected interests. They go so far as to concede that, in a proper case, persons in respondents' position might well recover damages for mental and emotional distress caused by the denial of procedural due process. . Petitioners' argument ■ is the more limited one that such injury cannot be presumed to occur, and that plaintiffs at least should be put to their proof on the issue, as plaintiffs are in most tort actions.

We agree with petitioners in this respect. As we have observed in another context, the doctrine of presumed damages in the common law of defamation per se “is an oddity of tort law, for it allows recovery of purportedly compensatory damages without evidence of actual loss.” Gertz v. Robert Welch, Inc., 418 U. S. 323, 349 (1974). The doctrine has been defended on the grounds that those forms of defamation that are actionable per se are virtually certain to cause serious injury to reputation, and that this kind of injury is extremely difficult to prove. See id., at 373, 376 (White, J., dissenting). 17 Moreover, statements that are defamatory per se by their very nature are likely to cause mental and emotional distress, as well as injury to reputation, so there arguably is little reason to require proof of this kind of injury either. 18 *263 But these considerations do not support respondents’ contention that damages should be presumed to flow from every deprivation of procedural due process.

First, it is not reasonable to assume that every departure from procedural due process, no matter what the circumstances or how minor, inherently is as likely to cause distress as the publication of defamation per se is to cause injury to reputation and distress. Where the deprivation of a protected interest is substantively justified but procedures are deficient in some respect, there may well be those who suffer no distress over the procedural irregularities. Indeed, in contrast to the immediately distressing effect of defamation per se, a person may not even know that procedures were deficient until he enlists the aid of counsel to challenge a perceived substantive deprivation.

Moreover, where a deprivation is justified but procedures are deficient, whatever distress a person feels may be attributable to the justified deprivation rather than to deficiencies in procedure. But as the Court of Appeals held, the injury caused by a justified deprivation, including distress, is not properly compensable under § 1983. 19 This ambiguity in causation, which is absent in the case of defamation per se, provides additional need for requiring the plaintiff to convince the trier of fact that he actually suffered distress because of the denial of procedural due process itelf.

Finally, we foresee no particular difficulty in producing evidence that mental and emotional distress actually was caused by the denial of procedural due process itself. Distress is a personal injury familiar to the law, customarily proved by *264 showing the nature and circumstances of the wrong and its effect on the plaintiff. 20 In sum, then, although mental and emotional distress caused by the denial of procedural due process itself is compensable under § 1983, we hold that neither the likelihood of such injury nor the difficulty of proving it is so great as to justify awarding compensatory damages without proof that such injury actually was caused.

D

The Court of Appeals believed, and respondents urge, that cases dealing with awards of damages for racial discrimination, the denial of voting rights, and the denial of Fourth Amendment rights support a presumption of damages where procedural due process is denied. 21 Many of the cases relied upon do not help respondents because they held or implied that some actual, if intangible, injury must be proved before compensatory damages may be recovered. Others simply did not address the issue. 22 More importantly, the elements and *265 prerequisites for recovery of damages appropriate to compensate injuries caused by the deprivation of one constitutional right are not necessarily appropriate to compensate injuries caused by the deprivation of another. As we have said, supra, at 258-259, these issues must be considered with reference to the nature of the interests protected by the particular constitutional right in question. For this reason, and without intimating an opinion as to their merits, we do- not deem the cases relied upon to be controlling.

*266 Ill

Even if respondents’ suspensions were justified, and even if they did not suffer any other actual injury, the fact remains that they were deprived of their right to procedural due process. “It is enough to invoke the procedural safeguards of the Fourteenth Amendment that a significant property interest is at stake, whatever the ultimate outcome of a hearing . . . .” Fuentes v. Shevin, 407 U. S., at 87; see Codd v. Velger, 429 U. S., at 632 (Stevens, J., dissenting); Coe v. Armour Fertilizer Works, 237 U. S. 413, 424 (1915).

Common-law courts traditionally have vindicated deprivations of certain “absolute” rights that are not shown to have caused actual injury through the award of a nominal sum of money. 23 By making the deprivation of such rights actionable for nominal damages without proof of actual injury, the law recognizes the importance to organized society that those rights be scrupulously observed; but at the same time, it remains true to the principle that substantial damages should be awarded only to compensate actual injury or, in the case of exemplary or punitive damages, to deter or punish malicious deprivations of rights.

Because the right to procedural due process is “absolute” in the sense that it does not depend upon the merits of a claimant’s substantive assertions, and because of the importance to organized society that procedural due process be observed, see Boddie v. Connecticut, 401 U. S. 371, 375 (1971) ; Anti-Fascist Committee v. McGrath, 341 U. S., at 171-172 (Frankfurter, J., concurring), we believe that the denial of procedural due process should be actionable for nominal damages without proof of actual injury. 24 We therefore hold that *267 if, upon remand, the District Court determines that respondents’ suspensions were justified, respondents nevertheless will be entitled to recover nominal damages not to exceed one dollar from petitioners. 25

The judgment of the Court of Appeals is reversed, and the case is remanded for further proceedings consistent with this opinion.

It is so ordered.

Justice Powell

delivered the opinion of the Court.

This case requires us to decide whether 42 U. S. C. § 1983 authorizes an award of compensatory damages based on the factfinder’s assessment of the value or importance of a substantive constitutional right.

M

Respondent Edward Stachura is a tenured teacher in the Memphis, Michigan, public schools. When the events that led to this case occurred, respondent taught seventh-grade life science, using a textbook that had been approved by the School Board. The textbook included a chapter on human reproduction. During the 1978-1979 school year, respondent spent six weeks on this chapter. As part of their instruction, students were shown pictures of respondent’s wife dur*301ing her pregnancy. Respondent also showed the students two films concerning human growth and sexuality. These films were provided by the County Health Department, and the Principal of respondent’s school had approved their use. Both films had been shown in past school years without incident.

After the showing of the pictures and the films, a number of parents complained to school officials about respondent’s teaching methods. These complaints, which appear to have been based largely on inaccurate rumors about the allegedly sexually explicit nature of the pictures and films, were discussed at an open School Board meeting held on April 23, 1979. Following the advice of the School Superintendent, respondent did not attend the meeting, during which a number of parents expressed the view that respondent should not be allowed to teach in the Memphis school system.1 The day after the meeting, respondent was suspended with pay. The School Board later confirmed the suspension, and notified respondent that an “administration evaluation” of his teaching methods was underway. No such evaluation was ever made. Respondent was reinstated the next fall, after filing this lawsuit.

Respondent sued the School District, the Board of Education, various Board members and school administrators, and two parents who had participated in the April 23 School Board meeting. The complaint alleged that respondent’s suspension deprived him of both liberty and property without due process of law and violated his First Amendment right to *302academic freedom. Respondent sought compensatory and punitive damages under 42 U. S. C. § 1983 for these constitutional violations.

At the close of trial on these claims, the District Court instructed the jury as to the law governing the asserted bases for liability. Turning to damages, the court instructed the jury that on finding liability it should award a sufficient amount to compensate respondent for the injury caused by petitioners’ unlawful actions:

“You should consider in this regard any lost earnings; loss of earning capacity; out-of-pocket expenses; and any mental anguish or emotional distress that you find the Plaintiff to have suffered as a result of conduct by the Defendants depriving him of his civil rights.” App. 94.

In addition to this instruction on the standard elements of compensatory damages, the court explained that punitive damages could be awarded, and described the standards governing punitive awards.2 Finally, at respondent’s request and over petitioners’ objection, the court charged that damages also could be awarded based on the value or importance of the constitutional rights that were violated:

“If you find that the Plaintiff has been deprived of a Constitutional right, you may award damages to compensate him for the deprivation. Damages for this type of injury are more difficult to measure than damages for a physical injury or injury to one’s property. There are no medical bills or other expenses by which you can judge how much compensation is appropriate. In one sense, no monetary value we place upon Constitutional rights can measure their importance in our society or compensate a citizen adequately for their deprivation. However, just because these rights are not capable of *303precise evaluation does not mean that an appropriate monetary amount should not be awarded.

“The precise value you place upon any Constitutional right which you find was denied to Plaintiff is within your discretion. You may wish to consider the importance of the right in our system of government, the role which this right has played in the history of our republic, [and] the significance of the right in the context of the activities which the Plaintiff was engaged in at the time of the violation of the right.” Id., at 96.

The jury found petitioners liable,3 and awarded a total of $275,000 in compensatory damages and $46,000 in punitive damages.4 The District Court entered judgment notwithstanding the verdict as to one of the defendants, reducing the total award to $266,750 in compensatory damages and $36,000 in punitive damages.

In an opinion devoted primarily to liability issues, the Court of Appeals for the Sixth Circuit affirmed, holding that respondent’s suspension had violated both procedural due process and the First Amendment. Stachura v. Truszkowski, 763 F. 2d 211 (1985). Responding to petitioners’ contention that the District Court improperly authorized damages based solely on the value of constitutional rights, the court noted only that “there was ample proof of actual injury to plaintiff Stachura both in his effective discharge . . . and by the damage to his reputation and to his professional career as a teacher. Contrary to the situation in Carey v. Piphus, 435 U. S. 247 (1978) . . . , there was proof from which the jury *304could have found, as it did, actual and important damages.” Id., at 214.

We granted certiorari limited to the question whether the Court of Appeals erred in affirming the damages award in the light of the District Court’s instructions that authorized not only compensatory and punitive damages, but also damages for the deprivation of “any constitutional right.”5 474 U. S. 918 (1985). We reverse, and remand for a new trial limited to the issue of compensatory damages.

l — H

Petitioners challenge the jury instructions authorizing damages for violation of constitutional rights on the ground that those instructions permitted the jury to award damages based on its own unguided estimation of the value of such rights.6 Respondent disagrees with this characterization of *305the jury instructions, contending that the compensatory damages instructions taken as a whole focused solely on respondent’s injury and not on the abstract value of the rights he asserted.

We believe petitioners more accurately characterize the instructions. The damages instructions were divided into three distinct segments: (i) compensatory damages for harm to respondent, (ii) punitive damages, and (iii) additional “compensat[ory]” damages for violations of constitutional rights. No sensible juror could read the third of these segments to modify the first.7 On the contrary, the damages instructions plainly authorized — in addition to punitive damages — two distinct types of “compensatory” damages: one based on respondent’s actual injury according to ordinary tort law standards, and another based on the “value” of certain rights. We therefore consider whether the latter category of damages was properly before the jury.

HH I — I h-H

A

We have repeatedly noted that 42 U. S. C. § 19838 creates “‘a species of tort liability’ in favor of persons who are deprived of ‘rights, privileges, or immunities secured’ to them *306by the Constitution.” Carey v. Piphus, 435 U. S. 247, 253 (1978), quoting Imbler v. Pachtman, 424 U. S. 409, 417 (1976). See also Smith v. Wade, 461 U. S. 30, 34 (1983); Newport v. Fact Concerts, Inc., 453 U. S. 247, 258-259 (1981). Accordingly, when § 1983 plaintiffs seek damages for violations of constitutional rights, the level of damages is ordinarily determined according to principles derived from the common law of torts. See Smith v. Wade, supra, at 34; Carey v. Piphus, supra, at 257-258; cf. Monroe v. Pape, 365 U. S. 167, 196, and n. 5 (1961) (Harlan, J., concurring).

Punitive damages aside,9 damages in tort cases are designed to provide “compensation for the injury caused to plaintiff by defendant’s breach of duty.” 2 F. Harper, F. James, & O. Gray, Law of Torts §25.1, p. 490 (2d ed. 1986) (emphasis in original), quoted in Carey v. Piphus, supra, at 255. See also Bivens v. Six Unknown Federal Narcotics Agents, 403 U. S. 388, 395, 397 (1971); id., at 408-409 (Har*307lan, J., concurring in judgment). To that end, compensatory-damages may include not only out-of-pocket loss and other monetary harms, but also such injuries as “impairment of reputation . . . , personal humiliation, and mental anguish and suffering.” Gertz v. Robert Welch, Inc., 418 U. S. 323, 350 (1974). See also Carey v. Piphus, supra, at 264 (mental and emotional distress constitute compensable injury in § 1983 cases). Deterrence is also an important purpose of this system, but it operates through the mechanisrn of damages that are compensatory — damages grounded in determinations of plaintiffs’ actual losses. E. g., 4 Harper, James, & Gray, supra, § 25.3 (discussing need for certainty in damages determinations); D. Dobbs, Law of Remedies § 3.1, pp. 135-136 (1973). Congress adopted this common-law system of recovery when it established liability for “constitutional torts.”10 Consequently, “the basic purpose” of § 1983 damages is “to compensate persons for injuries that are caused by the deprivation of constitutional rights.” Carey v. Piphus, 435 U. S., at 254 (emphasis added). See also id., at 257 (“damages awards under § 1983 should be governed by the principle of compensation”).

Carey v. Piphus represents a straightforward application of these principles. Carey involved a suit by a high school student suspended for smoking marijuana; the student claimed that he was denied procedural due process because he was suspended without an opportunity to respond to the charges against him. The Court of Appeals for the Seventh Circuit held that even if the suspension was justified, the student could recover substantial compensatory damages simply because of the insufficient procedures used to suspend him from school. We reversed, and held that the student could recover compensatory damages only if he proved actual injury caused by the denial of his constitutional rights. Id., at 264. We noted: “Rights, constitutional and otherwise, do *308not exist in a vacuum. Their purpose is to protect persons from injuries to particular interests . . . Id., at 254. Where no injury was present, no “compensatory” damages could be awarded.

The instructions at issue here cannot be squared with Carey, or with the principles of tort damages on which Carey and § 1983 are grounded. The jurors in this case were told that, in determining how much was necessary to “compensate [respondent] for the deprivation” of his constitutional rights, they should place a money value on the “rights” themselves by considering such factors as the particular right’s “importance ... in our system of government,” its role in American history, and its “significance ... in the context of the activities” in which respondent was engaged. App. 96. These factors focus, not on compensation for provable injury, but on the jury’s subjective perception of the importance of constitutional rights as an abstract matter. Carey establishes that such an approach is impermissible. The constitutional right transgressed in Carey — the right to due process of law — is central to our system of ordered liberty. See In re Gault, 387 U. S. 1, 20-21 (1967). We nevertheless held that no compensatory damages could be awarded for violation of that right absent proof of actual injury. Carey, 435 U. S., at 264. Carey thus makes clear that the abstract value of a constitutional right may not form the basis for § 1983 damages.11

*309Respondent nevertheless argues that Carey does not control here, because in this case a substantive constitutional right — respondent’s First Amendment right to academic freedom12 — was infringed. The argument misperceives our analysis in Carey. That case does not establish a two-tiered system of constitutional rights, with substantive rights afforded greater protection than “mere” procedural safeguards. We did acknowledge in Carey that “the elements and prerequisites for recovery of damages” might vary depending on the interests protected by the constitutional right at issue. Id., at 264-265. But we emphasized that, whatever the constitutional basis for § 1983 liability, such damages must always be designed “to compensate injuries caused by the [constitutional] deprivation.” Id., at 265 (emphasis added).13 See also Hobson v. Wilson, 237 U. S. App. D. C. 219, 277-279, 737 F. 2d 1, 59-61 (1984), cert. denied, 470 U. S. 1084 (1985); cf. Smith v. Wade, 461 U. S. 30 (1983). That conclusion simply leaves no room for noncompensatory *310damages measured by the jury’s perception of the abstract “importance” of a constitutional right.

Nor do we find such damages necessary to vindicate the constitutional rights that § 1983 protects. See n. 11, supra. Section 1983 presupposes that damages that compensate for actual harm ordinarily suffice to deter constitutional violations. Carey, supra, at 256-257 (“To the extent that Congress intended that awards under §1983 should deter the deprivation of constitutional rights, there is no evidence that it meant to establish a deterrent more formidable than that inherent in the award of compensatory damages”). Moreover, damages based on the “value” of constitutional rights are an unwieldy tool for ensuring compliance with the Constitution. History and tradition do not afford any sound guidance concerning the precise value that juries should place on constitutional protections. Accordingly, were such damages available, juries would be free to award arbitrary amounts without any evidentiary basis, or to use their unbounded discretion to punish unpopular defendants. Cf. Gertz, 418 U. S., at 350. Such damages would be too uncertain to be of any great value to plaintiffs, and would inject caprice into determinations of damages in §1983 cases. We therefore hold that damages based on the abstract “value” or “importance” of constitutional rights are not a permissible element of compensatory damages in such cases.

B

Respondent further argues that the challenged instructions authorized a form of “presumed” damages — a remedy that is both compensatory in nature and traditionally part of the range of tort law remedies. Alternatively, respondent argues that the erroneous instructions were at worst harmless error.

Neither argument has merit. Presumed damages are a substitute for ordinary compensatory damages, not a supplement for an award that fully compensates the alleged injury. When a plaintiff seeks compensation for an injury that is *311likely to have occurred but difficult to establish, some form of presumed damages may possibly be appropriate. See Carey, 435 U. S., at 262; cf. Dun & Bradstreet, Inc. v. Greenmoss Builders, 472 U. S. 749, 760-761 (1985) (opinion of Powell, J.); Gertz v. Robert Welch, Inc., supra, at 349. In those circumstances, presumed damages may roughly approximate the harm that the plaintiff suffered and thereby compensate for harms that may be impossible to measure. As we earlier explained, the instructions at issue in this case did not serve this purpose, but instead called on the jury to measure damages based on a subjective evaluation of the importance of particular constitutional values. Since such damages are wholly divorced from any compensatory purpose, they cannot be justified as presumed damages.14 *312Moreover, no rough substitute for compensatory damages was required in this case, since the jury was fully authorized to compensate respondent for both monetary and nonmone-tary harms caused by petitioners’ conduct.

Nor can we find that the erroneous instructions were harmless. See 28 U. S. C. § 2111; McDonough Power Equipment, Inc. v. Greenwood, 464 U. S. 548 (1984). When damages instructions are faulty and the verdict does not reveal the means by which the jury calculated damages, “[the] error in the charge is difficult, if not impossible, to correct without retrial, in light of the jury’s general verdict.” Newport v. Fact Concerts, Inc., 453 U. S., at 256, n. 12. The jury was authorized to award three categories of damages: (i) compensatory damages for injury to respondent, (ii) punitive damages, and (iii) damages based on the jury’s perception of the “importance” of two provisions of the Constitution. The submission of the third of these categories was error. Although the verdict specified an amount for punitive damages, it did not specify how much of the remaining damages was designed to compensate respondent for his injury and how much reflected the jury’s estimation of the value of the constitutional rights that were infringed. The effect of the erroneous instruction is therefore unknowable, although probably significant: the jury awarded respondent a very substantial amount of damages, none of which could have derived from any monetary loss.15 It is likely, although not certain, that a *313major part of these damages was intended to “compensate” respondent for the abstract “value” of his due process and First Amendment rights. For these reasons, the case must be remanded for a new trial on compensatory damages.

> hH

The judgment of the Court of Appeals is reversed, and the case is remanded for further proceedings consistent with this opinion.

It is so ordered.

Justice Brennan and Justice Stevens join the opinion of the Court and also join Justice Marshall’s opinion concurring in the judgment.

Justice Marshall,

with whom Justice Brennan, Justice Blackmun, and Justice Stevens join, concurring in the judgment.

I agree with the Court that this case must be remanded for a new trial on damages. Certain portions of the Court’s opinion, however, can be read to suggest that damages in § 1983 cases are necessarily limited to “out-of-pocket loss,” “other monetary harms,” and “such injuries as ‘impairment of reputation . . . , personal humiliation, and mental anguish and suffering.’” See ante, at 307. I do not understand the Court so to hold, and I write separately to emphasize that the violation of a constitutional right, in proper cases, may itself constitute a compensable injury.

The appropriate starting point of any analysis in this area is this Court’s opinion in Carey v. Piphus, 435 U. S. 247 (1978). In Carey, we recognized that “the basic purpose of a § 1983 damages award should be to compensate persons for injuries caused by the deprivation of constitutional rights.” Id., at 254; see ante, at 306-307. We explained, however, that application of that principle to concrete cases was not a *314simple matter. 435 U. S., at 257. “It is not clear,” we stated, “that common-law tort rules of damages will provide a complete solution to the damages issue in every § 1983 case.” Id., at 258. Rather, “the rules governing compensation for injuries caused by the deprivation of constitutional rights should be tailored to the interests protected by the particular right in question — just as the common-law rules of damages themselves were defined by the interests protected in various branches of tort law.” Id., at 259.

Applying those principles, we held in Carey that substantial damages should not be awarded where a plaintiff has been denied procedural due process but has made no further showing of compensable damage. We repeated, however, that “the elements and prerequisites for recovery of damages appropriate to compensate injuries caused by the deprivation of one constitutional right are not necessarily appropriate to compensate injuries caused by the deprivation of another.” Id., at 264-265. We referred to cases that support the award of substantial damages simply upon a showing that a plaintiff was wrongfully deprived of the right to vote, without requiring any further demonstration of damages. Id., at 264-265, n. 22.

Following Carey, the Courts of Appeals have recognized that invasions of constitutional rights sometimes cause injuries that cannot be redressed by a wooden application of common-law damages rules.* In Hobson v. Wilson, 237 U. S. App. D. C. 219, 275-281, 737 F. 2d 1, 57-63 (1984), cert. denied, 470 U. S. 1084 (1985), which the Court cites, ante, at 309, and n. 13, plaintiffs claimed that defendant Federal Bureau of Investigation agents had invaded their First *315Amendment rights to assemble for peaceable political protest, to associate with others to engage in political expression, and to speak on public issues free of unreasonable government interference. The District Court found that the defendants had succeeded in diverting plaintiffs from, and impeding them in, their protest activities. The Court of Appeals for the District of Columbia Circuit held that that injury to a First Amendment-protected interest could itself constitute compensable injury wholly apart from any “emotional distress, humiliation and personal indignity, emotional pain, embarassment, fear, anxiety and anguish” suffered by plaintiffs. 237 U. S. App. D. C., at 280, 737 F. 2d, at 62 (footnotes omitted). The court warned, however, that that injury could be compensated with substantial damages only to the extent that it was “reasonably quantifiable”; damages should not be based on “the so-called inherent value of the rights violated.” Ibid.

I believe that the Hobson court correctly stated the law. When a plaintiff is deprived, for example, of the opportunity to engage in a demonstration to express his political views, “[i]t is facile to suggest that no damage is done.” Dellums v. Powell, 184 U. S. App. D. C. 275, 303, 566 F. 2d 167, 195 (1977). Loss of such an opportunity constitutes loss of First Amendment rights “ fin their most pristine and classic form.’ ” Ibid., quoting Edwards v. South Carolina, 372 U. S. 229, 235 (1963). There is no reason why such an injury should not be compensable in damages. At the same time, however, the award must be proportional to the actual loss sustained.

The instructions given the jury in this case were improper because they did not require the jury to focus on the loss actually sustained by respondent. Rather, they invited the jury to base its award on speculation about “the importance of the right in our system of government” and “the role which this right has played in the history of our republic,” guided only by the admonition that “[i]n one sense, no monetary value we place on Constitutional rights can measure their im*316portance in our society or compensate a citizen adequately for their deprivation.” App. 96. These instructions invited the jury to speculate on matters wholly detached from the real injury occasioned respondent by the deprivation of the right. Further, the instructions might have led the jury to grant respondent damages based on the “abstract value” of the right to procedural due process — a course directly barred by our decision in Carey.

The Court therefore properly remands for a new trial on damages. I do not understand the Court, however, to hold that deprivations of constitutional rights can never themselves constitute compensable injuries. Such a rule would be inconsistent with the logic of Carey, and would defeat the purpose of § 1983 by denying compensation for genuine injuries caused by the deprivation of constitutional rights.

*249Justice Blackmun

delivered the opinion of the Court.

In Monell v. New York City Dept. of Social Services, 436 U. S. 658 (1978), this Court for the first time held that a local government was subject to suit as a “person” within the meaning of 42 U. S. C. § 1983. Aside from concluding that a municipal body was not wholly immune from civil liability, the Court had no occasion to explore the nature or scope of any particular municipal immunity under the statute. 436 U. S., at 701. The question presented by this case is whether a municipality may be held liable for punitive damages under § 1983.

I

A

Respondent Fact Concerts, Inc., is a Rhode Island corporation organized for the purpose of promoting musical concerts.1 In 1975, it received permission from the Rhode Island Depart*250ment of Natural Resources to present several summer concerts at Fort Adams, a state park located in the city of Newport. In securing approval for the final concerts, to be held August 30 and 31, respondent sought and obtained an entertainment license from petitioner city of Newport.2 Under their written contract, respondent retained control over the choice of performers and the type of music to be played while the city reserved the right to cancel the license without liability if “in the opinion of the City the interests of public safety demand.” App. 27.

Respondent engaged a number of well-known jazz music acts to perform during the final August concerts. Shortly before the dates specified, the group Blood, Sweat and Tears was hired as a replacement for a previously engaged performer who was unable to appear. Members of the Newport City Council, including the Mayor, became concerned that Blood, Sweat and Tears, which they characterized as a rock group rather than as a jazz band, would attract a rowdy and undesirable audience to Newport. 2 Record Appendix (R. A.) 265, 316-317, 325.3 Based on this concern, the Council attempted to have Blood, Sweat and Tears removed from the program.

On Monday, August 25, Mayor Donnelly informed respondent by telephone that he considered Blood, Sweat and Tears to be a rock group, and that they would not be permitted to perform because the city had experienced crowd disturbances at previous rock concerts. Id., at 195. Officials of respondent appeared before the City Council at a special meeting the next day, and explained that Blood, Sweat and Tears in fact were a jazz band that had performed at Carnegie Hall in New York City and at similar symphony hall facilities *251throughout the world. Speaking for the Council, the Mayor reiterated that the city did not condone rock festivals. Without attempting to investigate either the nature of the group’s music or the representations made by respondent, the Council voted to cancel the license for both days unless Blood, Sweat and Tears were removed from the program. Id., at 267-269. The vote received considerable publicity, and this adversely affected ticket sales. Id., at 248-G.

Later in the same week, respondent was informed by the City Solicitor that the Council had changed its position and would allow Blood, Sweat and Tears to perform if they did not play rock music. On Thursday, August 28, respondent agreed to attend a second special Council meeting the following day.

The second Council session convened on the afternoon of August 29, the day before the first scheduled performance. Mayor Donnelly informed the Council members that the city had two options — it could either allow Blood, Sweat and Tears to perform subject to the prohibition against rock music, or cancel the concert altogether. Although the City Solicitor advocated the first alternative and advised that cancellation would be unlawful, 3 R. A. 478, the Council did not offer the first option to respondent. Instead, one of the Council members inquired whether all provisions of the contract had been fulfilled. The City Manager, who had just returned from the concert site, reported that the wiring together of the spectator seats was not fully completed by 3 p. m., and that the auxiliary electric generator was not in place. Under the contract, respondent had agreed to fulfill these two conditions as part of the overall safety procedures. App. 28.4 The *252Council then voted to cancel the contract because respondent had not “lived up to all phases” of the agreement. 4 R. A. 10. The Council offered respondent a new contract for the same dates, specifically excluding Blood, Sweat and Tears. Respondent, however, indicated that it would take legal action if the original contract was not honored. 1 R. A. 96; 2 R. A. 202; 4 R. A; 11. After the meeting adjourned at 9:30 p.m., the decision to revoke respondent's license was broadcast extensively over the local media. 1 R. A. 97; 2 R. A. 204.

On Saturday morning, August 30, respondent obtained in state court a restraining order enjoining the Mayor, the City Council, and the city from interfering with the performance of the concerts. The 2-day event, including the appearance of Blood, Sweat and Tears, took place without incident. Fewer than half the available tickets were sold.

B

Respondent instituted the present action in the United States District Court for the District of Rhode Island, naming the city, its Mayor, and the six other Council members as defendants. Alleging, inter alia, that the license cancellation amounted to content-based censorship, and that its constitutional rights to free expression and due process had been violated under color of state law, respondent sought compensatory and punitive damages against the city and its officials under 42 U. S. C. § 1983 and under two pendent state-law counts, including tortious interference with contractual relationships. App. 8. At the conclusion of six days of trial, the District Court charged the jury with respect to the § 1983 and tortious interference counts. Included in its charge was *253an instruction, given without objection, that authorized the jury to award punitive damages against each defendant individually, “based on the degree of culpability of the individual defendant.” App. 62.5 The jury returned verdicts for respondent on both counts, awarding compensatory damages of $72,910 and punitive damages of $275,000; of the punitive damages, $75,000 was spread among the seven individual officials and $200,000 was awarded against the city.6

Petitioner moved for .a new trial, arguing that punitive damages cannot be awarded under § 1983 against a municipality, and that even if they can, the award was excessive.7 Because petitioner challenged the punitive damages instruction to which it had not objected at trial, the District Court noted that the challenge was untimely under Federal Rule of Civil Procedure 51. But the court was determined not to “rest its decision on this procedural ground alone.” App. to Pet. for Cert. B-3. Reasoning that “a careful resolution of this novel question is critical to a just verdict in this case,” *254id., at B-7, the court proceeded to consider petitioner’s substantive legal arguments on their merits.

The District Court recognized, ibid., that Monell had left undecided the question whether municipalities may be held liable for punitive damages. 436 U. S., at 701. The court observed, however, that punitive damages often had been awarded against individual officials in § 1983 actions, and it found no clear basis for distinguishing between individuals and municipalities in this regard. Emphasizing the general deterrent purpose served by punitive damages awards, the court reasoned that a municipality’s payment of such an award would focus taxpayer and voter attention upon the entity's malicious conduct, and that this in turn might promote accountability at the next election. App. to Pet. for Cert. B-9. Although noting that the burden imposed upon taxpaying citizens warranted judicial caution in this area, the court concluded that in appropriate circumstances municipalities could be held liable for punitive damages in a § 1983 action.8

The United States Court of Appeals for the First Circuit affirmed. 626 F. 2d 1060 (1980). That court noted, as an initial matter, that the challenge to the punitive damages award was flawed due to petitioner’s failure to object to the charge at trial. The court observed that such a failure should be overlooked “only where the error is plain and 'has seriously affected the fairness, integrity or public reputation of a judicial proceeding.’ ” Id., at 1067. The court found none of these factors present, because the law concerning municipal liability under § 1983 was in a state of flux, and no appellate decision had barred punitive damages awards against a municipality.

The Court of Appeals also expressed a belief that the *255challenged instruction might well not have been error at all. 626 F. 2d, at 1067. Citing its own prior holdings to the effect that punitive damages are available against § 1983 defendants, and this Court’s recent determination in Monell that a municipality is a “person” within the meaning of § 1983, the court identified the “distinct possibility that municipalities, like all other persons subject to suit under § 1983, may be liable for punitive damages in the proper circumstances.” 626 F. 2d, at 1067.

Because of the importance of the issue, we granted cer-tiorari. 449 U. S. 1060 (1980).

II

At the outset, respondent asserts that the punitive damages issue was not properly preserved for review before this Court. Brief for Respondents 7-9. In light of Rule 51’s uncompromising language9 and the policies of fairness and judicial efficiency incorporated therein, respondent claims that petitioner’s failure to object to the charge at trial should foreclose any further challenge to that instruction. The problem with respondent’s argument is that the District Court in the first instance declined to accept it. Although the punitive damages question perhaps could have been avoided simply by a reliance, under Rule 51, upon petitioner’s procedural default,10 the judge concluded that the interests of justice required careful consideration of this “novel question” of federal law.11 *256Because the District Court reached and fully adjudicated the merits, and the Court of Appeals did not disagree with that adjudication, no interests in fair and effective trial administration advanced by Rule 51 would be served if we refused now to reach the merits ourselves.12

Nor are we persuaded that our review should be limited to determining whether “plain error” has been committed, an exception to Rule 51 that is invoked on occasion by the Courts of Appeals absent timely objection in the trial court.13 No “right” to a specific standard of review exists in this setting, any more than a “right” to review existed at all once petitioner failed to except to the charge at trial. But given the special circumstances of this case, limiting our review to a restrictive “plain error” standard would be peculiarly inapt.

“Plain error” review under Rule 51 is suited to correcting obvious instances of injustice or misapplied law. A court’s interpretation of the contours of municipal liability under § 1983, as both courts below recognized, hardly could give rise to plain judicial error since those contours are currently in a state of evolving definition and uncertainty. See Owen v. City of Independence, 445 U. S. 622 (1980); Monell. See *257also Maine v. Thiboutot, 448 U. S. 1 (1980); Middlesex County Sewerage Authority v. National Sea Clammers Assn., ante, p. 1. We undertake review here in order to resolve one element of the uncertainty, that is, the availability of punitive damages, and it would scarcely be appropriate or just to confine our review to determining whether any error that might exist is sufficiently egregious to qualify under Rule 51. The very novelty of the legal issue at stake counsels uncon-stricted review.

In addition to being novel, the punitive damages question is important and appears likely to recur in § 1983 litigation against municipalities.14 And here the question was squarely presented and decided on a complete trial record by the court of first resort, was argued by both sides to the Court of Appeals, and has been fully briefed before this Court. In light of all these factors, we conclude that restricting our review to the plain-error standard would serve neither to promote the interests of justice nor to advance efficient judicial administration.15 We therefore turn to the merits of petitioner’s claim.16

*258Ill

It. is by now well settled that the tort liability created by § 1983 cannot be understood in a historical vacuum. In the Civil Rights Act of 1871, Congress created a federal remedy against a person who, acting under color of state law, deprives another of constitutional rights. See Monroe v. Pape, 365 U. S. 167, 172 (1961). Congress, however, expressed no intention to do away with the immunities afforded state officials at common law, and the Court consistently has declined to construe the general language of § 1983 17 as automatically abolishing such traditional immunities by implication. Procunier v. Navarette, 434 U. S. 555, 561 (1978); Imbler v. Pachtman, 424 U. S. 409, 417 (1976); Pierson v. Ray, 386 U. S. 547, 554-555 (1967); Tenney v. Brandhove, 341 U. S. 367, 376 (1951). Instead, the Court has recognized immunities of varying scope applicable to different officials sued under the statute.18 One important assumption underlying the Court’s decisions in this area is that members of the 42d Congress were familiar with common-law principles, including defenses previously recognized in ordinary tort litigation, and that they likely intended these common-law principles to obtain, absent specific provisions to the contrary.

At the same time, the Court’s willingness to recognize certain -traditional immunities as affirmative defenses has not led it to conclude that Congress incorporated all immunities exist*259ing at common law. See Scheuer v. Rhodes, 416 U. S. 232, 243 (1974). Indeed, because the 1871 Act was designed to expose state and local officials to a new form of liability, it would defeat the promise of the statute to recognize any preexisting immunity without determining both the policies that it serves and its compatibility with the purposes of § 1983. See Imbler v. Pachtman, 424 U. S., at 424; id., at 434 (opinion concurring in judgment); Owen v. City of Independence, 445 U. S., at 638. Only after careful inquiry into considerations of both history and policy has the Court construed § 1983 to incorporate a particular immunity defense.

Since Monell was decided three years ago, the Court has applied this two-part approach when scrutinizing a claim of immunity proffered by a municipality. In Owen v. City of Independence, the Court held that neither history nor policy supported a construction of § 1983 that would allow a municipality to assert the good faith of its officers or agents as a defense to liability for damages. 445 U. S., at 638, 657. Owen, however, concerned only compensatory damages, and petitioner contends that with respect to a municipality’s liability for punitive damages, an examination of the common-law background and policy considerations yields a very different result.

A

By the time Congress enacted what is now § 1983, the immunity of a municipal corporation from punitive damages at common law was not open to serious question. It was generally understood by 1871 that a municipality, like a private corporation, was to be treated as a natural person subject to suit for a wide range of tortious activity,19 but this understand*260ing did not extend to the award of punitive or exemplary damages. Indeed, the courts that had considered the issue prior to 1871 were virtually unanimous in denying such damages against a municipal corporation. E. g., Woodman v. Nottingham, 49 N. H. 387 (1870); City of Chicago v. Langlass, 52 Ill, 256 (1869); City Council of Montgomery v. Gilmer & Taylor, 33 Ala. 116 (1858); Order of Hermits of St. Augustine v. County of Philadelphia, 4 Clark 120, Brightly N. P. 116 (Pa. 1847); McGary v. President & Council of the City of Lafayette, 12 Rob. 668, 674 (La. 1846).20 Judicial disinclination to award punitive damages against a municipality has persisted to the present day in the vast majority of jurisdictions.21 See generally 18 E. McQuillin, Municipal Corporations § 53.18a (3d rev. ed. 1977); P. Burdick, Law of Torts 245-246 (4th ed. *2611926); 4 J. Dillon, Law of Municipal Corporations § 1712 (5th ed. 1911); G. Field, Law of Damages §80 (1876).

The language of the opinions themselves is instructive as to the reasons behind this common-law tradition. In McGary, for example, the Louisiana Supreme Court refused to allow punitive damages against the city of Lafayette despite the malicious acts of its municipal officers, who had violated an injunction by ordering the demolition of plaintiff’s house. Reasoning that the officials’ malice should not be attributed to the taxpaying citizens of the community, the court explained its holding:

“Those who violate the laws of their country, disregard the authority of courts of justice, and wantonly inflict injuries, certainly become thereby obnoxious to vindictive damages. These, however, can never be allowed against the innocent. Those which the plaintiff has recovered in the present case . . . , being evidently vindictive, cannot, in our opinion, be sanctioned by this court, as they are to be borne by widows, orphans, aged men and women, and strangers, who, admitting that they must repair the injury inflicted by the Mayor on the plaintiff, cannot be bound beyond that amount, which will be sufficient for her indemnification.” 12 Rob., at 677.

Similarly, in Hunt v. City of Boonville, 65 Mo. 620 (1877), the Missouri Supreme Court held that a municipality could not be found liable for treble damages under a trespass statute, notwithstanding the statute’s authorization of such damages against “any person.” After noting the existence of “respectable authority” to the effect that municipal corporations “can not, as such, do a criminal act or a willful and malicious wrong and they cannot therefore be made liable for exemplary damages,” id., at 624, the court continued:

“[T]he relation which the officers of a municipal corporation sustain toward the citizens thereof for whom they act, is not in all respects identical with that existing be*262tween the stockholders of a private corporation and their agents; and there is not the same reason for holding municipal corporations, engaged in the performance of acts for the public benefit, liable for the willful or malicious acts of its officers, as there is in the case of private corporations.” Id., at 625.

Of particular relevance to our current inquiry is Order of Hermits of St. Augustine v. County of Philadelphia, supra, which involved a Pennsylvania statute that authorized property owners within the county to bring damages actions against it for the destruction of their property by mob violence.22 The court observed that the “persons” against whom the statute authorized recovery included the county corporation, and it held that plaintiffs were entitled to compensatory damages as part of the county’s duty to make reparation to its citizens for injuries sustained as a result of lawless violence. While noting that punitive damages would have been available against the. rioters themselves, the court nonetheless held that such exemplary damages were not recoverable against the county.

The rationale of these decisions was reiterated in numerous other common-law jurisdictions. E. g., Wilson v. City of Wheeling, 19 W. Va. 323, 350 (1882) (“The city is not a spoliator and should not be visited by vindictive or punitive damages”); City of Chicago v. Langlass, 52 Ill., at 259 (“But in fixing the compensation the jury have no right to give vindictive or punitive damages, against a municipal corporation. Against such a body they should only be compensatory, and not by way of punishment”); City Council of Montgomery v. Gilmer & Taylor, 33 Ala., at 132 (“The [municipal] corporation can not, upon any principle known *263to us, be responsible for the malice of its officers towards the plaintiffs”). In general, courts viewed punitive damages as contrary to sound public policy, because such awards would burden the very taxpayers and citizens for whose benefit the wrongdoer was being chastised. The courts readily distinguished between liability to compensate for injuries inflicted by a municipality’s officers and agents, and vindictive damages appropriate as punishment for the bad-faith conduct of those same officers and agents. Compensation was an obligation properly shared by the municipality itself, whereas punishment properly applied only to the actual wrongdoers. The courts thus protected the public from unjust punishment, and the municipalities from undue fiscal constraints.23

Given that municipal immunity from punitive damages was well established at common law by 1871, we proceed on the familiar assumption that “Congress would have specifically so provided had it wished to abolish the doctrine.” Pierson v. Ray, 386 U. S., at 555. Nothing in the legislative debates suggests that, in enacting § 1 of the Civil Rights Act, *264the 42d Congress intended any such abolition. Indeed, the limited legislative history relevant to this issue suggests the opposite.

Because there was virtually no debate on § 1 of the Act, the Court has looked to Congress’ treatment of the amendment to the Act introduced by Senator Sherman as indicative of congressional attitudes toward the nature and scope of municipal liability. Monell, 436 U. S., at 692, n. 57.24 Initially, it is significant that the Sherman amendment as proposed contemplated the award of no more than compensatory damages for injuries inflicted by mob violence. The amendment would not have exposed municipal governments to punitive damages; rather, it proposed that municipalities “shall be liable to pay full compensation to the person or persons damnified” by mob violence. Globe, at 749, 765 (emphasis added).25 *265That the exclusion of punitive damages was no oversight was confirmed by Representative Butler, one of the amendment’s chief supporters, when he responded to a critical inquiry on the floor of the House:

“The invalidity of the gentleman’s argument is that he looks upon [the amendment] as a punishment for the county. Now, we do not look upon it as a punishment at all. It is a mutual insurance. We are there a community, and if there is any wrong done by our community, or by the inhabitants of our community, we will indemnify the injured party for that wrong . . . .” Id., at 792.

We doubt that a Congress having no intention of permitting punitive awards against municipalities in the explicit context of the Sherman amendment would have meant to expose municipal bodies to such novel liability sub silentio under § 1 of the Act.

Notwithstanding the compensatory focus of the amendment, its proposed extension of municipal liability met substantial resistance in Congress, resulting in its defeat on two separate occasions.26 In addition to the constitutional reservations broached by legislators, which the Court has discussed at some length in Monell, 436 U. S., at 669-683, Members of both Chambers also expressed more practical objections. Notably, supporters as well as opponents of § 1 voiced concern that this extension of public liability might place an unmanageable financial burden on local governments.27 Legislators *266also expressed apprehension that innocent taxpayers would be unfairly punished for the deeds of persons over whom they had neither knowledge nor control.28 Admittedly, both these objections were raised with particular reference to the threat of the expansive municipal liability embodied in the Sherman amendment. The two concerns are not without relevance to the present inquiry, however, in that they reflect policy considerations similar to those relied upon by the common-law courts in rejecting punitive damages awards. We see no reason to believe that Congress’ opposition to punishing innocent taxpayers and bankrupting local governments would have been less applicable with regard to the novel specter of punitive damages against municipalities.

B

Finding no evidence that Congress intended to disturb the settled common-law immunity, we now must determine whether considerations of public policy dictate a contrary result. In doing so, we examine the objectives underlying punitive damages in general, and their relationship to the goals of § 1983.

Punitive damages by definition are not intended to compensate the injured party, but rather to punish the tortfeasor *267whose wrongful action was intentional or malicious, and to deter him and others from similar extreme conduct. See Restatement (Second) of Torts §908 (1979); W. Prosser, Law of Torts 9-10 (4th ed. 1971). Regarding retribution, it remains true that an award of punitive damages against a municipality “punishes” only the taxpayers, who took no part in the commission of the tort. These damages are assessed over and above the amount necessary to compensate the injured party. Thus, there is no question here of equitably distributing the losses resulting from official misconduct. Cf. Owen v. City of Independence, 445 U. S., at 657. Indeed, punitive damages imposed on a municipality are in effect a windfall to a fully compensated plaintiff, and are likely accompanied by an increase in taxes or a reduction of public services for the citizens footing the bill. Neither reason nor justice suggests that such retribution should be visited upon the shoulders of blameless or unknowing taxpayers.29

Under ordinary principles of retribution, it is the wrongdoer himself who is made to suffer for his unlawful conduct. If a government official acts knowingly and maliciously to deprive others of their civil rights, he may become the appropriate object of the community’s vindictive sentiments. See generally Silver v. Cormier, 529 F. 2d 161, 163 (CA10 1976); Bucher v. Krause, 200 F. 2d 576, 586-588 (CA7 1952), cert, denied, 345 U. S. 997 (1953). A municipality, however, can have no malice independent of the malice of its officials. Damages awarded for punitive purposes, therefore, are not sensibly assessed against the governmental entity itself.

To the extent that the purposes of § 1983 have any bearing on this punitive rationale, they do not alter our analysis. The Court previously has indicated that punitive damages *268might be awarded in appropriate circumstances in order to punish violations of constitutional rights, Carey v. Piphus, 435 U. S. 247, 257, n. 11 (1978), but it never has suggested that punishment is as prominent a purpose under the statute as are compensation and deterrence. See, e. g., Owen v. City of Independence, 445 U. S., at 651; Robertson v. Wegmann, 436 U. S. 584, 590-591 (1978); Carey v. Piphus, 435 U. S., at 256-257. Whatever its weight, the retributive purpose is not significantly advanced, if it is advanced at all, by exposing municipalities to punitive damages.

The other major objective of punitive damages awards is to prevent future misconduct. Respondent argues vigorously that deterrence is a primary purpose of § 1983, and that because punitive awards against municipalities for the malicious conduct of their policymaking officials will induce voters to condemn official misconduct through the electoral process, the threat of such awards will deter future constitutional violations. Brief for Respondents 9-11. Respondent is correct in asserting that the deterrence of future abuses of power by persons acting under color of state law is an important purpose of § 1983. Owen v. City of Independence, 445 U. S., at 651; Robertson v. Wegmann, 436 U. S., at 591. It is in this context that the Court’s prior statements contemplating punitive damages “in ‘a proper’ § 1983 action” should be understood. Carlson v. Green, 446 U. S. 14, 22 (1980); Carey v. Piphus, 435 U. S., at 257, n. 11. For several reasons, however, we conclude that the deterrence rationale of § 1983 does not justify making punitive damages available against municipalities.

First, it is far from clear that municipal officials, including those at the policymaking level, would be deterred from wrongdoing by the knowledge that large punitive awards could be assessed based on the wealth of their municipality. Indemnification may not be available to the municipality under local law, and even if it were, officials likely will not be able themselves to pay such sizable awards. Thus, assum*269ing, arguendo, that the responsible official is not impervious to shame and humiliation, the impact on the individual tort-feasor of this deterrence in the air is at best uncertain.

There also is no reason to suppose that corrective action, such as the discharge of offending officials who were appointed and the public excoriation of those who were elected, will not occur unless punitive damages are awarded against the municipality. The Court recently observed in a related context: “The more reasonable assumption is that responsible superiors are motivated not only by concern for the public fisc but also by concern for the Government’s integrity.” Carlson v. Green, 446 U. S., at 21. This assumption is no less applicable to the electorate at large. And if additional protection is needed, the compensatory damages that are available against a municipality may themselves induce the public to vote the wrongdoers out of office.

Moreover, there is available a more effective means of deterrence. By allowing juries and courts to assess punitive damages in appropriate circumstances against the offending official, based on his personal financial resources, the statute directly advances the public’s interest in preventing repeated constitutional deprivations.30 In our view, this provides sufficient protection against the prospect that a public official may *270commit recurrent constitutional violations by reason of his office. The Court previously has found, with respect to such violations, that a damages remedy recoverable against individuals is more effective as a deterrent than the threat of damages against a government employer. Carlson v. Green, 446 U. S., at 21. We see no reason to depart from that conclusion here, especially since the imposition of additional penalties would most likely fall upon the citizen-taxpayer.

Finally, although the benefits associated with awarding punitive damages against municipalities under § 1983 are of doubtful character, the costs may be very real. In light of the Court’s decision last Term in Maine v. Thiboutot, 448 U. S. 1 (1980), the § 1983 damages remedy may now be available for violations of federal statutory as well as constitutional law. But cf. Middlesex County Sewerage Authority v. National Sea Clammers Assn., ante, p. 1. Under this expanded liability, municipalities and other units of state and local government face the possibility of having to assure compensation for persons harmed by abuses of governmental authority covering a large range of activity in everyday life. To add the burden of exposure for the malicious conduct of individual government employees may create a serious risk to the financial integrity of these governmental entities.

The Court has remarked elsewhere on the broad discretion traditionally accorded to juries in assessing the amount of punitive damages. Electrical Workers v. Foust, 442 U. S. 42, 50-51 (1979); Gertz v. Robert Welch, Inc., 418 U. S. 323, 349-350 (1974). Because evidence of a tortfeasor’s wealth is traditionally admissible as a measure of the amount of punitive damages that should be awarded,31 the unlimited taxing power of a municipality may have a prejudicial impact on the jury, in effect encouraging it to impose a sizable award. The impact of such a windfall recovery is likely to be both un*271predictable and, at times, substantial, and we are sensitive to the possible strain on local treasuries and therefore on services available to the public at large.32 Absent a compelling reason for approving such an award, not present here, we deem it unwise to inflict the risk.

IV

In sum, we find that considerations of history and policy do not support exposing a municipality to punitive damages for the bad-faith actions of its officials. Because absolute immunity from such damages obtained at common law and was undisturbed by the 42d Congress, and because that immunity is compatible with both the purposes of § 1983 and general.principles of public policy, we hold that a municipality is immune from punitive damages under 42 U. S. C. § 1983. Accordingly, the judgment of the Court of Appeals is vacated, and the case is remanded for further proceedings consistent with this opinion.

It is so ordered.

Justice Brennan,

with whom Justice Marshall and Justice Stevens join, dissenting.

The Court today considers and decides a challenge to the District Court’s jury instructions, even though petitioners failed to object to the instructions in a timely manner, as required by Rule 51 of the Federal Rules of Civil Procedure. Because this departure from Rule 51 is unprecedented and unwarranted, I respectfully dissent.

Respondents filed suit against petitioners in Federal District Court under 42 U. S. C. § 1983, alleging violations of their *272First Amendment rights. In their complaint and amended complaint, respondents prayed for punitive damages, as well as other relief. App. 11, 12, 13, 24, 25, 26. Respondents submitted a pretrial memorandum on the issue of punitive damages and, during trial, submitted an additional memorandum on the availability of punitive damages against a municipal corporation, in response to the court’s request to both parties. Brief in Opposition 8. At the close of the evidence, the court instructed the jury explicitly and in detail that it could impose punitive damages against petitioners if they had acted maliciously, wantonly, or oppressively. App. 57-58. After giving the instruction, the Court summoned the attorneys to the side bar, inviting objections or suggestions concerning the instructions. Record Appendix (R. A.) 591-A to 591-B. For reasons not revealed in the record, counsel for petitioners expressly declined to make any such objection or suggestion.1 Id., at 591-B. The jury returned a verdict in favor of respondents, and awarded substantial punitive damages against each of the petitioners, including the city of Newport.

Petitioners moved for judgment notwithstanding the verdict, and for a new trial, arguing, inter alia, that punitive damages may not be imposed against a municipality under § 1983. The court denied the motion, stating:

“None of these legal arguments were ever raised at trial. In fact, the defendants failed to request that any of their current legal interpretations be inserted into the jury instructions and never objected to any aspect of that charge before or after the jury retired. . . . Therefore, defendants’ untimely objections are not the proper basis for this post-trial motion.” App. to Pet. for Cert. B-2 to B-3 (citing Fed. Rule Civ. Proc. 51).

Petitioners’ failure to object to the punitive damages instruc*273tion thus precluded them from raising the issue on post-trial motions. Not content to “rest its decision on this procedural ground alone,” id., at B-3 (emphasis added), however, the court also held, in the alternative, that its punitive damages instruction was correct on the merits. Id., at B-7 to B-10.

On appeal to the Court of Appeals for the First Circuit, the court stated that petitioners’ allegation of error in the punitive damages instruction

“is flawed by the failure to object to the charge at trial. See Fed. R. Civ. P. 51. We may overlook a failure of this nature, but only where the error is plain and 'has seriously affected the fairness, integrity or public reputation of a judicial proceeding.’ ” 626 F. 2d 1060, 1067 (1980), quoting Morris v. Travisono, 528 F. 2d 856, 859 (CA1 1976) (footnote and citation omitted).

The Court of Appeals then briefly canvassed the relevant precedents, stated that the law concerning punitive damages against municipalities under § 1983 is in a “state of flux,” 626 F. 2d, at 1067, and concluded: “[W]e would be hard-pressed to say that the trial judge’s punitive damages instruction was plain error. Nor is this a case containing such 'peculiar circumstances [to warrant, noticing error] to prevent a clear miscarriage of justice.’ ” Id., at 1067-1068, quoting Nimrod v. Sylvester, 369 F. 2d 870, 873 (CA1 1966) (citation omitted; brackets in original).

Respondents argue before this Court that the decision of the Court of Appeals should be affirmed, because petitioners failed to object to the punitive damages instruction.2 They *274rely on Federal Rule of Civil Procedure 51, which states in relevant part: “No party may assign as error the giving or the failure to give an instruction unless he objects thereto before the jury retires to consider its verdict.”

Rule 51 could not be expressed more clearly. Cases too numerous to list have held that failure to object to proposed jury instructions in a timely manner in accordance with Rule 51 precludes appellate review.3 Rule 51 serves an important function in ensuring orderly judicial administration and fairness to the parties. The trial judge is thereby informed in precise terms of any objections to proposed instructions, and thus is given “an opportunity upon second thought, and before it is too late, to correct any inadvertent or erroneous failure to charge.” Marshall v. Nugent, 222 F. 2d 604, 615 (CA1 1955). Moreover, the Rule prevents litigants from making the tactical decision not to object to instructions at trial in order to preserve a ground for appeal. In light of the significant purposes and “uncompromising language,” ante, at 255, of Rule 51, courts should not depart lightly from its strictures.

Nevertheless, like other procedural rules, Rule 51 is susceptible to flexible interpretation when strictly necessary to *275avoid a clear miscarriage of justice. Cf. Wood v. Georgia, 450 U. S. 261, 265, n. 5 (1981); Carlson v. Green, 446 U. S. 14, 17, n. 2 (1980); Hormel v. Helvering, 312 U. S. 552, 557 (1941).4 Accordingly, the Courts of Appeals have developed a “plain error” doctrine to deal with certain unchallenged jury instructions so contrary to law as to be manifestly unjust. Whatever the proper scope of such a doctrine,5 courts and commentators uniformly agree that it should be applied only in exceptional circumstances. As the Court of Appeals for the First Circuit has noted: “ ‘If there is to be a plain error exception to Rule 51 at all, it should be confined to the exceptional case where the error has seriously affected the fairness, integrity, or public reputation of judicial proceedings.’ ” Morris v. Travisono, supra, at 859, quoting 9 C. Wright & A. Miller, Federal Practice and Procedure § 2558, p. 675 (1971). This was the standard applied by the Court of Appeals below. 626 F. 2d, at 1067.

The Court states that the “problem with” respondents’ argument that petitioners are barred from raising the punitive damages issue “is that the District Court in the first in*276stance declined to accept it.” Ante, at 255. But the District Court did not reject respondents’ argument; on the contrary, it expressly held that petitioners’ objections to the jury instructions were “untimely” under Rule 51, and therefore were “not the proper basis” for post-trial challenge. App. to Pet. for Cert. B-3. Its prudential decision to discuss the merits as well does not detract from this holding.6 As the Court of Appeals held, this procedural ground is sufficient to compel affirmance in the absence of a finding of plain error constituting manifest injustice. Petitioners themselves admit that the punitive damages question may be reviewed only under a plain-error standard. Brief for Petitioners 27.

The Court today frankly admits that the instruction was not plain error, noting that the governing principles of law are “currently in a state of evolving definition and uncertainty.” Ante, at 256. Nevertheless, it vacates the Court of Appeals’ judgment. Such a vacating necessarily implies that the Court of Appeals’ treatment of the procedural question was in error, but the Court provides not a hint as to what standard the Court of Appeals should have applied.7 Indeed, the Court *277does not even state in so many words that the Court of Appeals erred, much less explain why.

The Court does assert that under the “special circumstances of this case” it would be “peculiarly inapt” to confine our review to the plain-error standard employed below. It explains that the issue in this case is “novel,” and that it “appears likely to recur.” Ante, at 256, 257. But most of the issues before this Court are novel and likely to recur: that is why they are considered worthy of certiorari. And to the extent issues are novel, it behooves us to grant certiorari in eases where there has been full consideration of the issues by the courts below, rather than cursory treatment under a plain-error standard.

The Court also suggests that this case is somehow “special” because the issue “was squarely presented and decided on a complete record by the court of first resort, was argued by both sides to the Court of Appeals, and has been fully briefed before this Court.” Ante, at 257. But these factors are present whenever the District Court reconsiders unchallenged jury instructions on the merits as an alternative holding, the *278Court of Appeals affirms on a plain-error standard, and this Court grants certiorari. See n. 6, supra. In short, I see the circumstances of this case as anything but “special.”

Applying settled principles, I conclude that the Court of Appeals was correct to affirm the District Court in this case. The jury instruction, as the Court admits, did not constitute “plain error.” Moreover, as the Court of Appeals held, failure to review the instruction would not cause a clear miscarriage of justice, any more than would failure to review any other unchallenged jury instruction. There is no reason to treat punitive damages instructions differently from other instructions for Rule 51 purposes. See Whiting v. Jackson State University, 616 F. 2d 116, 126-127 (CA5 1980) (no timely objection having been made, court’s failure to give punitive damages instruction upheld except in exceptional cases); Mid-America Food Service, Inc. v. ARA Services, Inc., 578 F. 2d 691 (CA8 1978) (no timely objection having been made, punitive damages instruction upheld in absence of plain error). Nor is the city of Newport entitled to special treatment by virtue of its governmental status. Cf. Morris v. Travisono, 528 F. 2d, at 859 (failure of state correctional officers in § 1983 suit to object to jury instructions not excused, even though the instructions directed the jury to apply a harsher constitutional standard than had been established by precedent).

Indeed, I consider this a peculiarly inapt case to disregard petitioners’ procedural default. There would be no injustice whatsoever in adhering to the Rule in this case. Petitioners were given clear notice that punitive damages would be an issue in the case; the jury instructions were unambiguous; petitioners had ample opportunity to object; they failed to do so, without offering any reason or excuse.8 Whether their *279default was negligent or tactical, they have no cause now to complain. If these petitioners’ default is to be excused, whose should not? If Rule 51 is to be disregarded in this case, when should it be enforced?

I dissent.

Justice White

delivered the opinion of the Court.

The issue here is whether respondent Lyons satisfied the prerequisites for seeking injunctive relief in the Federal District Court.

I

This case began on February 7, 1977, when respondent, Adolph Lyons, filed a complaint for damages, injunction, and declaratory relief in the United States District Court for the Central District of California. The defendants were the City of Los Angeles and four of its police officers. The complaint alleged that on October 6, 1976, at 2 a. m., Lyons was stopped by the defendant officers for a traffic or vehicle code violation and that although Lyons offered no resistance or threat whatsoever, the Officers, without provocation or justification, seized Lyons and applied a “chokehold”1 — either *98the “bar arm control” hold or the “carotid-artery control” hold or both — rendering him unconscious and causing damage to his larynx. Counts I through IV of the complaint sought damages against the officers and the City. Count V, with which we are principally concerned here, sought a preliminary and permanent injunction against the City barring the use of the control holds. That count alleged that the City’s police officers, “pursuant to the authorization, instruction and encouragement of Defendant City of Los Angeles, regularly and routinely apply these choke holds in innumerable situations where they are not threatened by the use of any deadly force whatsoever,” that numerous persons have been injured as the result of the application of the chokeholds, that Lyons and others similarly situated are threatened with irreparable injury in the form of bodily injury and loss of life, and that Lyons “justifiably fears that any contact he has with Los Angeles Police officers may result in his being choked and strangled to death without provocation, justification or other legal excuse.” Lyons alleged the threatened impairment of rights protected by the First, Fourth, Eighth, and Fourteenth Amendments. Injunctive relief was sought against the use of the control holds “except in situations where the proposed victim of said control reasonably appears to be threatening the immediate use of deadly force.” Count VI sought declaratory relief against the City, i. e., a judgment that use of the chokeholds absent the threat of immediate use of deadly force is a per se violation of various constitutional rights.

The District Court, by order, granted the City’s motion for partial judgment on the pleadings and entered judgment for *99the City on Counts V and VI.2 The Court of Appeals reversed the judgment for the City on Counts V and VI, holding over the City’s objection that despite our decisions in O’Shea v. Littleton, 414 U. S. 488 (1974), and Rizzo v. Goode, 423 U. S. 362 (1976), Lyons had standing to seek relief against the application of the chokeholds. Lyons v. City of Los Angeles, 615 F. 2d 1243 (1980). The Court of Appeals held that there was a sufficient likelihood that Lyons would again be stopped and subjected to the unlawful use of force to constitute a case or controversy and to warrant the issuance of an injunction, if the injunction was otherwise authorized. We denied certiorari. 449 U. S. 934 (1980).

On remand, Lyons applied for a preliminary injunction. Lyons pressed only the Count V claim at this point. See n. 6, infra. The motion was heard on affidavits, depositions, and government records. The District Court found that Lyons had been stopped for a traffic infringement and that without provocation or legal justification the officers involved had applied a “Department-authorized chokehold which resulted in injuries to the plaintiff.” The court further found that the department authorizes the use of the holds in situations where no one is threatened by death or grievous bodily harm, that officers are insufficiently trained, that the use of the holds involves a high risk of injury or death as then employed, and that their continued use in situations where neither death nor serious bodily injury is threatened “is unconscionable in a civilized society.” The court concluded that such use violated Lyons’ substantive due process rights under the Fourteenth Amendment. A preliminary injunc*100tion was entered enjoining “the use of both the carotid artery and bar arm holds under circumstances which do not threaten death or serious bodily injury.” An improved training program and regular reporting and recordkeeping were also ordered.3 The Court of Appeals affirmed in a brief per curiam opinion stating that the District Court had not abused its discretion in entering a preliminary injunction. 656 F. 2d 417 (1981). We granted certiorari, 455 U. S. 937 (1982), and now reverse.

II

Since our grant of certiorari, circumstances pertinent to the case have changed. Originally, Lyons’ complaint alleged that at least two deaths had occurred as a result of the application of chokeholds by the police. His first amended complaint alleged that 10 chokehold-related deaths had occurred. By May 1982, there had been five more such deaths. On May 6,1982, the Chief of Police in Los Angeles prohibited the use of the bar-arm chokehold in any circumstances. A few days later, on May 12,1982, the Board of Police Commissioners imposed a 6-month moratorium on the use of the carotid-artery chokehold except under circumstances where deadly force is authorized.4

*101Based on these events, on June 3, 1982, the City filed in this Court a memorandum suggesting a question of mootness, reciting the facts but arguing that the case was not moot. Lyons in turn filed a motion to dismiss the writ of certiorari as improvidently granted. We denied that motion but reserved the question of mootness for later consideration. 457 U. S. 1115 (1982).

In his brief and at oral argument, Lyons has reasserted his position that in light of changed conditions, an injunctive decree is now unnecessary because he is no longer subject to a threat of injury. He urges that the preliminary injunction should be vacated. The City, on the other hand, while acknowledging that subsequent events have significantly changed the posture of this case, again asserts that the case is not moot because the moratorium is not permanent and may be lifted at any time.

We agree with the City that the case is not moot, since the moratorium by its terms is not permanent. Intervening events have not “irrevocably eradicated the effects of the alleged violation.” County of Los Angeles v. Davis, 440 U. S. 625, 631 (1979). We nevertheless hold, for another reason, that the federal courts are without jurisdiction to entertain Lyons’ claim for injunctive relief.

HH I — I l-H

It goes without saying that those who seek to invoke the jurisdiction of the federal courts must satisfy the threshold requirement imposed by Art. Ill of the Constitution by alleging an actual case or controversy. Flast v. Cohen, 392 U. S. 83, 94-101 (1968); Jenkins v. McKeithen, 395 U. S. 411, 421-425 (1969) (opinion of Marshall, J.). Plaintiffs must demonstrate a “personal stake in the outcome” in order to “assure that concrete adverseness which sharpens the presentation of issues” necessary for the proper resolution of constitutional questions. Baker v. Carr, 369 U. S. 186, 204 (1962). Abstract injury is not enough. The plaintiff must *102show that he “has sustained or is immediately in danger of sustaining some direct injury” as the result of the challenged official conduct and the injury or threat of injury must be both “real and immediate,” not “conjectural” or “hypothetical.” See, e. g., Golden v. Zwickler, 394 U. S. 103, 109-110 (1969); Public Workers v. Mitchell, 330 U. S. 75, 89-91 (1947); Maryland Casualty Co. v. Pacific Coal & Oil Co., 312 U. S. 270, 273 (1941); Massachusetts v. Mellon, 262 U. S. 447, 488 (1923).

In O’Shea v. Littleton, 414 U. S. 488 (1974), we dealt with a case brought by a class of plaintiffs claiming that they had been subjected to discriminatory enforcement of the criminal law. Among other things, a county magistrate and judge were accused of discriminatory conduct in various respects, such as sentencing members of plaintiff’s class more harshly than other defendants. The Court of Appeals reversed the dismissal of the suit- by the District Court, ruling that if the allegations were proved, an appropriate injunction could be entered.

We reversed for failure of the complaint to allege a case or controversy. Id., at 493. Although it was claimed in that case that particular members of the plaintiff class had actually suffered from the alleged unconstitutional practices, we observed that “[p]ast exposure to illegal conduct does not in itself show a present case or controversy regarding injunctive relief ... if unaccompanied by any continuing, present adverse effects.” Id., at 495-496. Past wrongs were evidence bearing on “whether there is a real and immediate threat of repeated injury.” Id., at 496. But the prospect of future injury rested “on the likelihood that [plaintiffs] will again be arrested for and charged with violations of the criminal law and will again be subjected to bond proceedings, trial, or sentencing before petitioners.” Ibid. The most that could be said for plaintiffs’ standing was “that if [plaintiffs] proceed to violate an unchallenged law and if they are charged, held to answer, and tried in any proceedings before petitioners, they will be subjected to the discriminatory prac*103tices that petitioners are alleged to have followed.” Id., at 497. We could not find a case or controversy in those circumstances: the threat to the plaintiffs was not “sufficiently real and immediate to show an existing controversy simply because they anticipate violating lawful criminal statutes and being tried for their offenses. ...” Id., at 496. It was to be assumed that “[plaintiffs] will conduct their activities within the law and so avoid prosecution and conviction as well as exposure to the challenged course of conduct said to be followed by petitioners.” Id., at 497.

We further observed that case-or-controversy considerations “obviously shade into those determining whether the complaint states a sound basis for equitable relief,” id., at 499, and went on to hold that even if the complaint presented an existing case or controversy, an adequate basis for equitable relief against petitioners had not been demonstrated:

“[Plaintiffs] have failed, moreover, to establish the basic requisites of the issuance of equitable relief in these circumstances — the likelihood of substantial and immediate irreparable injury, and the inadequacy of remedies at law. We have already canvassed the necessarily conjectural nature of the threatened injury to which [plaintiffs] are allegedly subjected. And if any of the [plaintiffs] are ever prosecuted and face trial, or if they are illegally sentenced, there are available state and federal procedures which could provide relief from the wrongful conduct alleged.” Id., at 502.

Another relevant decision for present purposes is Rizzo v. Goode, 423 U. S. 362 (1976), a case in which plaintiffs alleged widespread illegal and unconstitutional police conduct aimed at minority citizens and against city residents in general. The Court reiterated the holding in O’Shea that past wrongs do not in themselves amount to that real and immediate threat of injury necessary to make out a case or controversy. The claim of injury rested upon “what one of a small, unnamed minority of policemen might do to them in the future *104because of that unknown policeman’s perception” of departmental procedures. 423 U. S., at 372. This hypothesis was “even more attenuated than those allegations of future injury found insufficient in O’Shea to warrant [the] invocation of federal jurisdiction.” Ibid. The Court also held that plaintiffs’ showing at trial of a relatively few instances of violations by individual police officers, without any showing of a deliberate policy on behalf of the named defendants, did not provide a basis for equitable relief.

Golden v. Zwickler, 394 U. S. 103 (1969), a case arising in an analogous situation, is directly apposite. Zwickler sought a declaratory judgment that a New York statute prohibiting anonymous handbills directly pertaining to election campaigns was unconstitutional. Although Zwickler had once been convicted under the statute,5 his sole concern related to a Congressman who had left the House of Representatives for a place on the Supreme Court of New York and who would not likely be a candidate again. A unanimous Court held that because it was “most unlikely” that Zwickler would again be subject to the statute, no case or controversy of “‘sufficient immediacy and reality’” was present to allow a declaratory judgment. Id., at 109. Just as Zwickler’s assertion that the former Congressman could be a candidate for Congress again was “hardly a substitute for evidence that this is a prospect of ‘immediacy and reality,’” ibid., Lyons’ assertion that he may again be subject to an illegal chokehold does not create the actual controversy that must exist for a declaratory judgment to be entered.

We note also our per curiam opinion in Ashcroft v. Mattis, 431 U. S. 171 (1977). There, the father of a boy who had been killed by the police sought damages and a declaration that the Missouri statute which authorized police officers to use deadly force in apprehending a person who committed a felony was unconstitutional. Plaintiff alleged that he had an*105other son, who “‘if ever arrested or brought under an attempt at arrest on suspicion of a felony, might flee or give the appearance of fleeing, and would therefore be in danger of being killed by these defendants or other police officers ....’” Id., at 172, n. 2. We ruled that “[s]uch speculation is insufficient to establish the existence of a present, live controversy.” Id., at 173, n. 2.

IV

No extension of O Shea and Rizzo is necessary to hold that respondent Lyons has failed to demonstrate a case or controversy with the City that would justify the equitable relief sought.6 Lyons’ standing to seek the injunction requested depended on whether he was likely to suffer future injury from the use of the chokeholds by police officers. Count V of the complaint alleged the traffic stop and choking incident five months before. That Lyons may have been illegally choked by the police on October 6, 1976, while presumably affording Lyons standing to claim damages against the individual officers and perhaps against the City, does nothing to establish a real and immediate threat that he would again be stopped for a traffic violation, or for any other offense, by an officer or officers who would illegally choke him into unconsciousness without any provocation or resistance on his part. The additional allegation in the complaint that the police in Los Angeles routinely apply chokeholds in situations where they are not threatened by the use of deadly force falls far short of the allegations that would be necessary to establish a case or controversy between these parties.

In order to establish an actual controversy in this case, Lyons would have had not only to allege that he would have *106another encounter with the police but also to make the incredible assertion either (1) that all police officers in Los An-geles always choke any citizen with whom they happen to have an encounter, whether for the purpose of arrest, issuing a citation, or for questioning, or (2) that the City ordered or authorized police officers to act in such manner. Although Count V alleged that the City authorized the use of the control holds in situations where deadly force was not threatened, it did not indicate why Lyons might be realistically threatened by police officers who acted within the strictures of the City’s policy. If, for example, chokeholds were authorized to be used only to counter resistance to an arrest by a suspect, or to thwart an effort to escape, any future threat to Lyons from the City’s policy or from the conduct of police officers would be no more real than the possibility that he would again have an encounter with the police and that either he would illegally resist arrest or detention or the officers would disobey their instructions and again render him unconscious without any provocation.7

*107Under O’Shea and Rizzo, these allegations were an insufficient basis to provide a federal court with jurisdiction to entertain Count V of the complaint.8 This was apparently the conclusion of the District Court in dismissing Lyons’ claim for injunctive relief. Although the District Court acted without opinion or findings, the Court of Appeals interpreted its action as based on lack of standing, i. e., that under O’Shea and Rizzo, Lyons must be held to have made an “insufficient showing that the police were likely to do this to the plaintiff again.” 615 F. 2d, at 1246. For several reasons — each of. them infirm, in our view — the Court of Appeals thought reliance on O’Shea and Rizzo was misplaced and reversed the District Court.

First, the Court of Appeals thought that Lyons was more immediately threatened than the plaintiffs in those cases since, according to the Court of Appeals, Lyons need only *108be stopped for a minor traffic violation to be subject to the strangleholds. But even assuming that Lyons would again be stopped for a traffic or other violation in the reasonably near future, it is untenable to assert, and the complaint made no such allegation, that strangleholds are applied by the Los Angeles police to every citizen who is stopped or arrested regardless of the conduct of the person stopped. We cannot agree that the “odds,” 615 F. 2d, at 1247, that Lyons would not only again be stopped for a traffic violation but would also be subjected to a chokehold without any provocation whatsoever are sufficient to make out a federal case for equitable relief. We note that five months elapsed between October 6, 1976, and the filing of the complaint, yet there was no allegation of further unfortunate encounters between Lyons and the police.

Of course, it may be that among the countless encounters between the police and the citizens of a great city such as Los Angeles, there will be certain instances in which strangleholds will be illegally applied and injury and death unconstitutionally inflicted on the victim. As we have said, however, it is no more than conjecture to suggest that in every instance of a traffic stop, arrest, or other encounter between the police and a citizen, the police will act unconstitutionally and inflict injury without provocation or legal excuse. And it is surely no more than speculation to assert either that Lyons himself will again be involved in one of those unfortunate instances, or that he will be arrested in the future and provoke the use of a chokehold by resisting arrest, attempting to escape, or threatening deadly force or serious bodily injury.

Second, the Court of Appeals viewed O’Shea and Rizzo as cases in which the plaintiffs sought “massive structural” relief against the local law enforcement systems and therefore that the holdings in those cases were inapposite to cases such as this where the plaintiff, according to the Court of Appeals, seeks to enjoin only an “established,” “sanctioned” police practice assertedly violative of constitutional rights. O’Shea and Rizzo, however, cannot be so easily confined to their *109facts. If Lyons has made no showing that he is realistically threatened by a repetition of his experience of October 1976, then he has not met the requirements for seeking an injunction in a federal court, whether the injunction contemplates intrusive structural relief or the cessation of a discrete practice.

The Court of Appeals also asserted that Lyons “had a live and active claim” against the City “if only for a period of a few seconds” while the stranglehold was being applied to him and that for two reasons the claim had not become moot so as to disentitle Lyons to injunctive relief: First, because under normal rules of equity, a case does not become moot merely because the complained of conduct has ceased; and second, because Lyons’ claim is “capable of repetition but evading review” and therefore should be heard. We agree that Lyons had a live controversy with the City. Indeed, he still has a claim for damages against the City that appears to meet all Art. Ill requirements. Nevertheless, the issue here is not whether that claim has become moot but whether Lyons meets the preconditions for asserting an injunctive claim in a federal forum. The equitable doctrine that cessation of the challenged conduct does not bar an injunction is of little help in this respect, for Lyons’ lack of standing does not rest on the termination of the police practice but on the speculative nature of his claim that he will again experience injury as the result of that practice even if continued.

The rule that a claim does not become moot where it is capable of repetition, yet evades review, is likewise inapposite. Lyons’ claim that he was illegally strangled remains to be litigated in his suit for damages; in no sense does that claim “evade” review. Furthermore, the capable-of-repetition doctrine applies only in exceptional situations, and generally only where the named plaintiff can make a reasonable showing that he will again be subjected to the alleged illegality. DeFunis v. Odegaard, 416 U. S. 312, 319 (1974). As we have indicated, Lyons has not made this demonstration.

*110The record and findings made on remand do not improve Lyons’ position with respect to standing. The District Court, having been reversed, did not expressly address Lyons’ standing to seek injunctive relief, although the City was careful to preserve its position on this question. There was no finding that Lyons faced a real and immediate threat of again being illegally choked. The City’s policy was described as authorizing the use of the strangleholds “under circumstances where no one is threatened with death or grievous bodily harm.” That policy was not further described, but the record before the court contained the department’s existing policy with respect to the employment of chokeholds. Nothing in that policy, contained in a Police Department manual, suggests that the chokeholds, or other kinds of force for that matter, are authorized absent some resistance or other provocation by the arrestee or other suspect.9 On the contrary, police officers were instructed to use chokeholds only when lesser degrees of force do not suffice and then only “to gain control of a suspect who is violently resisting the officer or trying to escape.” App. 230.

Our conclusion is that the Court of Appeals failed to heed O’Shea, Rizzo, and other relevant authority, and that the District Court was quite right in dismissing Count V.

*111V

Lyons fares no better if it be assumed that his pending damages suit affords him Art. Ill standing to seek an injunction as a remedy for the claim arising out of the October 1976 events. The equitable remedy is unavailable absent a showing of irreparable injury, a requirement that cannot be met where there is no showing of any real or immediate threat that the plaintiff will be wronged again — a “likelihood of substantial and immediate irreparable injury.” O’Shea v. Littleton, 414 U. S., at 502. The speculative nature of Lyons’ claim of future injury requires a finding that this prerequisite of equitable relief has not been fulfilled.

Nor will the injury that Lyons allegedly suffered in 1976 go unrecompensed; for that injury, he has an adequate remedy at law. Contrary to the view of the Court of Appeals, it is not at all “difficult” under our holding “to see how anyone can ever challenge police or similar administrative practices.” 615 F. 2d, at 1250. The legality of the violence to which Lyons claims he was once subjected is at issue in his suit for damages and can be determined there. -

Absent a sufficient likelihood that he will again be wronged in a similar way, Lyons is no more entitled to an injunction than any other citizen of Los Angeles; and a federal court may not entertain a claim by .any or all citizens who no more than assert that certain practices of law enforcement officers are unconstitutional. Cf. Warth v. Seldin, 422 U. S. 490 (1975); Schlesinger v. Reservists to Stop the War, 418 U. S. 208 (1974); United States v. Richardson, 418 U. S. 166 (1974). This is not to suggest that such undifferentiated claims should not be taken seriously by local authorities. Indeed, the interest of an alert and interested citizen is an essential element of an effective and fair government, whether on the local, state, or national level.10 A federal court, how*112ever, is not the proper forum to press such claims unless the requirements for entry and the prerequisites for injunctive relief are satisfied.

We decline the invitation to slight the preconditions for equitable relief; for as we have held, recognition of the need for a proper balance between state and federal authority counsels restraint in the issuance of injunctions against state officers engaged in the administration of the States’ criminal laws in the absence of irreparable injury which is both great and immediate. O’Shea, supra, at 499; Younger v. Harris, 401 U. S. 37, 46 (1971). Mitchum v. Foster, 407 U. S. 225 (1972), held that suits brought under 42 U. S. C. § 1983 are exempt from the flat ban against the issuance of injunctions directed at state-court proceedings, 28 U. S. C. §2283. But this holding did not displace the normal principles of equity, comity, and federalism that should inform the judgment of federal courts when asked to oversee state law enforcement authorities. In exercising their equitable powers federal courts must recognize “[t]he special delicacy of the adjustment to be preserved between federal equitable power and State administration of its own law.” Stefanelli v. Minard, 342 U. S. 117, 120 (1951); O’Shea v. Littleton, supra, at 500. See also Rizzo v. Goode, 423 U. S., at 380; Cleary v. Bolger, 371 U. S. 392 (1963); Wilson v. Schnettler, 365 U. S. 381 (1961); Pugach v. Dollinger, 365 U. S. 458 (1961). The Court of Appeals failed to apply these factors properly and therefore erred in finding that the District Court had not abused its discretion in entering an injunction in this case.

As we noted in O’Shea, 414 U. S., at 503, withholding in-junctive relief does not mean that the “federal law will exer*113cise no deterrent effect in these circumstances.” If Lyons has suffered an injury barred by the Federal Constitution, he has a remedy for damages under § 1983. Furthermore, those who deliberately deprive a citizen of his constitutional rights risk conviction under the federal criminal laws. Ibid.

Beyond these considerations the state courts need not impose the same standing or remedial requirements that govern federal-court proceedings. The individual States may permit their courts to use injunctions to oversee the conduct of law enforcement authorities on a continuing basis. But this is not the role of a federal court, absent far more justification than Lyons has proffered in this case.

The judgment of the Court of Appeals is accordingly

Reversed.

Justice Marshall,

with whom Justice Brennan, Justice Blackmun, and Justice Stevens join, dissenting.

The District Court found that the city of Los Angeles authorizes its police officers to apply life-threatening choke-holds to citizens who pose no threat of violence, arid that respondent, Adolph Lyons, was subjected to such a chokehold. The Court today holds that a federal court is without power to enjoin the enforcement of the city’s policy, no matter how flagrantly unconstitutional it may be. Since no one can show that he will be choked in the future, no one — not even a person who, like Lyons, has almost been choked to death — has standing to challenge the continuation of the policy. The city is free to continue the policy indefinitely as long as it is willing to pay damages for the injuries and deaths that result. I dissent from this unprecedented and unwarranted approach to standing.

There is plainly a “case or controversy” concerning the constitutionality of the city’s chokehold policy. The constitutionality of that policy is directly implicated by Lyons’ claim for damages against the city. The complaint clearly alleges *114that the officer who choked Lyons was carrying out an official policy, and a municipality is liable under 42 U. S. C. § 1983 for the conduct of its employees only if they acted pursuant to such a policy. Monell v. New York City Dept. of Social Services, 436 U. S. 658, 694 (1978). Lyons therefore has standing to challenge the city’s chokehold policy and to obtain whatever relief a court may ultimately deem appropriate. None of our prior decisions suggests that his requests for particular forms of relief raise any additional issues concerning his standing. Standing has always depended on whether a plaintiff has a “personal stake in the outcome of the controversy,” Baker v. Carr, 369 U. S. 186, 204 (1962), not on the “precise nature of the relief sought.” Jenkins v. McKeithen, 395 U. S. 411, 423 (1969) (opinion of Marshall, J., joined by Warren, C. J., and Brennan, J.).

I

A

Respondent Adolph Lyons is a 24-year-old Negro male who resides in Los Angeles. According to the uncontradicted evidence in the record,1 at about 2 a. m. on October 6, 1976, Lyons was pulled over to the curb by two officers of the Los Angeles Police Department (LAPD) for a traffic infraction because one of his taillights was burned out. The officers greeted him with drawn revolvers as he exited from his car. Lyons was told to face his car and spread his legs. He did so. He was then ordered to clasp his hands and put them on top of his head. He again complied. After one of the officers completed a patdown search, Lyons dropped his hands, *115but was ordered to place them back above his head, and one of the officers grabbed Lyons’ hands and slammed them onto his head. Lyons complained about the pain caused by the ring of keys he was holding in his hand. Within 5 to 10 seconds, the officer began to choke Lyons by applying a forearm against his throat.' As Lyons struggled for air, the officer handcuffed him, but continued to apply the chokehold until he blacked out. When Lyons regained consciousness, he was lying face down on the ground, choking, gasping for air, and spitting up blood and dirt. He had urinated and defecated. He was issued a traffic citation and released.

On February 7, 1977, Lyons commenced this action under 42 U. S. C. § 1983 against the individual officers and the city, alleging violations of his rights under' the Fourth, Eighth, and Fourteenth Amendments to the Constitution and seeking damages and declaratory and injunctive relief. He claimed that he was subjected to a chokehold without justification and that defendant officers were “carrying out the official policies, customs and practices of the Los Angeles Police Department and the City of Los Angeles.” Count II, ¶ 13.2 These allegations were included or incorporated in each of the Counts in which the city was named as a defendant. See Counts II through VI. Lyons alleged that the city authorizes the use of chokeholds “in innumerable situations where [the police] are not threatened by the use of any deadly force whatsoever.” Count V, ¶ 22.

B

Although the city instructs its officers that use of a choke-hold does not constitute deadly force, since 1975 no less than 16 persons have died following the use of a chokehold by *116an LAPD police officer. Twelve have been Negro males.3 The evidence submitted to the District Court4 established that for many years it has been the official policy of the city to permit police officers to employ chokeholds in a variety of situations where they face no threat of violence. In reported “altercations” between LAPD officers and citizens the choke-holds are used more frequently than any other means of physical restraint.5 Between February 1975 and July 1980, LAPD officers applied chokeholds on at least 975 occasions, which represented more than three-quarters of the reported altercations.6

It is undisputed that chokeholds pose a high and unpredictable risk of serious injury or death. Chokeholds are intended to bring a subject under control by causing pain and rendering him unconscious. Depending on the position of the officer’s arm and the force applied, the victim’s voluntary *117or involuntary reaction, and his state of health, an officer may inadvertently crush the victim’s larynx, trachea, or hyoid. The result may be death caused by either cardiac arrest or asphyxiation.7 An LAPD officer described the reaction of a person to being choked as “do[ing] the chicken,” *118Exh. 44, p. 93, in reference apparently to the reactions of a chicken when its neck is wrung. The victim experiences extreme pain. His face turns blue as he is deprived of oxygen, he goes into spasmodic convulsions, his eyes roll back, his body wriggles, his feet kick up and down, and his arms move about wildly.

Although there has been no occasion to determine the precise contours of the city’s chokehold policy, the evidence submitted to the District Court provides some indications. LAPD Training Officer Terry Speer testified that an officer is authorized to deploy a chokehold whenever he “feels that there’s about to be a bodily attack made on him.” App. 381 (emphasis added). A training bulletin states that “[c]ontrol holds . . . allow officers to subdue any resistance by the suspects.” Exh. 47, p. 1 (emphasis added). In the proceedings below the city characterized its own policy as authorizing the use of chokeholds “ ‘to gain control of a suspect who is violently resisting the officer or trying to escape,’” to “subdue any resistance by the suspects,”8 and to permit an officer, “where . . . resisted, but not necessarily threatened with serious bodily harm or death, ... to subdue a suspect who forcibly resists an officer.” (Emphasis added.)9

The training given LAPD officers provides additional revealing evidence of the city’s chokehold policy. Officer *119Speer testified that in instructing officers concerning the use of force, the LAPD does not distinguish between felony and misdemeanor suspects. App. 379. Moreover, the officers are taught to maintain the chokehold until the suspect goes limp, id., at 387; App. to Pet. for Cert. 51a, despite substantial evidence that the application of a chokehold invariably induces a “flight or flee” syndrome, producing an involuntary struggle by the victim which can easily be misinterpreted by the officer as willful resistance that must be overcome by prolonging the chokehold and increasing the force applied. See n. 7, supra. In addition, officers are instructed that the chokeholds can be safely deployed for up to three or four minutes. App. 387-388; App. to Pet. for Cert. 48. Robert Jarvis, the city’s expert who has taught at the Los Angeles Police Academy for the past 12 years, admitted that officers are never told that the bar-arm control can cause death if applied for just two seconds. App. 388. Of the nine deaths for which evidence was submitted to the District Court, the average duration of the choke where specified was approximately 40 seconds.

C

In determining the appropriateness of a preliminary injunction, the District Court recognized that the city’s policy is subject to the constraints imposed by the Due Process Clause of the Fourteenth Amendment. The court found that “[d]uring the course of this confrontation, said officers, without provocation or legal justification, applied a Department-authorized chokehold which resulted in injuries to plaintiff.” (Emphasis added.) The court found that the “City of Los Angeles and the Department authorize the use of these holds under circumstances where no one is threatened by death or grievous bodily harm.” The court concluded that the use of the chokeholds constitutes “deadly force,” and that the city may not constitutionally authorize the use of such force “in situations where death or serious bodily harm is not threatened.” On the basis of this conclusion, the District Court en*120tered a preliminary injunction enjoining “the use of both the carotid-artery and bar arm holds under circumstances which do not threaten death or serious bodily injury.”10 As the Court of Appeals noted, “[a]ll the trial judge has done, so far, is to tell the city that its police officers may not apply life threatening strangleholds to persons stopped in routine police work unless the application of such force is necessary to prevent serious bodily harm to an officer.” 656 F. 2d 417, 418 (1981).

II

At the outset it is important to emphasize that Lyons’ entitlement to injunctive relief and his entitlement to an award of damages both depend upon whether he can show that the city’s chokehold policy violates the Constitution. An indispensable prerequisite of municipal liability under 42 U. S. C. §1983 is proof that the conduct complained of is attributable to an unconstitutional official policy or custom. Polk County v. Dodson, 454 U. S. 312, 326 (1981); Monell v. New York City Dept. of Social Services, 436 U. S., at 694. It is not enough for a § 1983 plaintiff to show that the employees or agents of a municipality have violated or will violate the Constitution, for a municipality will not be held liable solely on a theory of respondeat superior. See Monell, supra, at 694.

The Court errs in suggesting that Lyons’ prayer for injunc-tive relief in Count V of his first amended complaint concerns a policy that was not responsible for his injuries and that therefore could not support an award of damages. Ante, at 106-107, n. 7. Paragraph 8 of the complaint alleges that Lyons was choked “without provocation, legal justification or ex*121cuse.” Paragraph 13 expressly alleges that “[t]he Defendant Officers were carrying out the official policies, customs and practices of the Los Angeles Police Department and the City of Los Angeles,” and that “by virtue thereof, defendant City is liable for the actions” of the officers. (Emphasis added.) These allegations are incorporated in each of the Counts against the city, including Count V.

There is no basis for the Court’s assertion that Lyons has failed to allege “that the City either orders or authorizes application of the chokeholds where there is no resistance or other provocation.” Ante, at 106, n. 7. I am completely at a loss to understand how paragraphs 8 and 13 can be deemed insufficient to allege that the city’s policy authorizes the use of chokeholds without provocation. The Court apparently finds Lyons’ complaint wanting because, although it alleges that he was choked without provocation and that the officers acted pursuant to an official policy, it fails to allege in haec verba that the city’s policy authorizes the choking of suspects without provocation. I am aware of no case decided since the abolition of the old common-law forms of action, and the Court cites none, that in any way supports this crabbed construction of the complaint. A federal court is capable of concluding for itself that two plus two equals four.11

The Court also errs in asserting that even if the complaint sufficiently alleges that the city’s policy authorizes the use of chokeholds without provocation, such an allegation is in any event “belied by the record made on the application for preliminary injunction.” Ibid. This conclusion flatly contradicts the District Court’s express factual finding, which was left undisturbed by the Court of Appeals, that the officers applied a “Department-authorized chokehold which resulted in *122injuries to plaintiff.” (Emphasis added.) The city does not contend that this factual finding is clearly erroneous.12

In sum, it is absolutely clear that Lyons’ requests for damages and for injunctive relief call into question the constitutionality of the city’s policy concerning the use of chokeholds. If he does not show that that policy is unconstitutional, he will be no more entitled to damages than to an injunction.

III

Since Lyons’ claim for damages plainly gives him standing, and since the success of that claim depends upon a demonstration that the city’s chokehold policy is unconstitutional, it is beyond dispute that Lyons has properly invoked the District Court’s authority to adjudicate the constitutionality of the city’s chokehold policy. The disputé concerning the constitutionality of that policy plainly presents a “case or controversy” under Art. III. The Court nevertheless holds that a federal court has no power under Art. Ill to adjudicate Lyons’ request, in the same lawsuit, for injunctive relief with respect to that very policy. This anomalous result is not supported either by precedent or by the fundamental concern underlying the standing requirement. Moreover, by fragmenting a single claim into multiple claims for particular types of relief and requiring a separate showing of standing for each form of relief, the decision today departs from this *123Court’s traditional conception of standing and of the remedial powers of the federal courts.

A

It is simply disingenuous for the Court to assert that its decision requires “[n]o extension” of O’Shea v. Littleton, 414 U. S. 488 (1974), and Rizzo v. Goode, 423 U. S. 362 (1976). Ante, at 105. In contrast to this case O’Shea and Rizzo involved disputes focusing solely on the threat of future injury which the plaintiffs in those cases alleged they faced. In O’Shea the plaintiffs did not allege past injury and did not seek compensatory relief.13 In Rizzo, the plaintiffs sought only declaratory and injunctive relief and alleged past instances of police misconduct only in an attempt to establish the substantiality of the threat of future injury. There was similarly no claim for damages based on past injuries in Ashcroft v. Mattis, 431 U. S. 171 (1977), or Golden v. Zwickler, 394 U. S. 103 (1969),14 on which the Court also relies.

*124These decisions do not support the Court’s holding today. As the Court recognized in O’Shea, standing under Art. III is established by an allegation of “‘threatened or actual injury.’” 414 U. S., at 493, quoting Linda R. S. v. Richard D., 410 U. S. 614, 617 (1973) (emphasis added). See also 414 U. S., at 493, n. 2. Because the plaintiffs in O’Shea, Rizzo, Mattis, and Zwickler did not seek to redress past injury, their standing to sue depended entirely on the risk of future injury they faced. Apart from the desire to eliminate the possibility of future injury, the plaintiffs in those cases had no other personal stake in the outcome of the controversies.

By contrast, Lyons’ request for prospective relief is coupled with his claim for damages based on past injury. In addition to the risk that he will be subjected to a chokehold in the future, Lyons has suffered past injury.15 Because he has a live claim for damages, he need not rely solely on the threat of future injury to establish his personal stake in the outcome of the controversy.16 In the cases relied on by the majority, *125the Court simply had no occasion to decide whether a plaintiff who has standing to litigate a dispute must clear a separate standing hurdle with respect to each form of relief sought.17

B

The Court’s decision likewise finds no support in the fundamental policy underlying the Art. Ill standing requirement — the concern that a federal court not decide a legal issue if the plaintiff lacks a sufficient “personal stake in the outcome of the controversy as to assure that concrete adverseness which sharpens the presentation of issues upon which the court so largely depends for illumination of difficult . . . questions.” Baker v. Carr, 369 U. S., at 204. As this Court stated in Flast v. Cohen, 392 U. S. 83, 101 (1968), “the question of standing is related only to whether the dispute sought to be adjudicated will be presented in an adversary context and in a form historically viewed as capable of judicial resolution.” See also Valley Forge Christian College v. *126Americans United for Separation of Church and State, 454 U. S. 464, 472 (1982) (standing requirement ensures that “the legal questions presented to the court will be resolved, not in the rarified atmosphere of a debating society, but in a concrete factual context conducive to a realistic appreciation of the consequences of judicial action”).

Because Lyons has a claim for damages against the city, and because he cannot prevail on that claim unless he demonstrates that the city’s chokehold policy violates the Constitution, his personal stake in the outcome of the controversy adequately assures an adversary presentation of his challenge to the constitutionality of the policy.18 Moreover, the resolution of this challenge will be largely dispositive of his requests for declaratory and injunctive relief. No doubt the requests for injunctive relief may raise additional questions. But these questions involve familiar issues relating to the appropriateness of particular forms of relief, and have never been thought to implicate a litigant’s standing to sue. The denial of standing separately to seek injunctive relief therefore cannot be justified by the basic concern underlying the Art. Ill standing requirement.19

*127C

By fragmenting the standing inquiry and imposing a separate standing hurdle with respect to each form of relief sought, the decision today departs significantly from this Court’s traditional conception of the standing requirement and of the remedial powers of the federal courts. We have never required more than that a plaintiff have standing to litigate a claim. Whether he will be entitled to obtain particular forms of relief should he prevail has never been understood to be an issue of standing. In determining whether a plaintiff has standing, we have always focused on his personal stake in the outcome of the controversy, not on the issues sought to be litigated, Flast v. Cohen, supra, at 99, or the “precise nature of the relief sought.” Jenkins v. McKeithen, 395 U. S., at 423 (opinion of Marshall, J., joined by Warren, C. J., and Brennan, J.).

*1281

Our cases uniformly state that the touchstone of the Art. Ill standing requirement is the plaintiff’s personal stake in the underlying dispute, not in the particular types of relief sought. Once a plaintiff establishes a personal stake in a dispute, he has done all that is necessary to “invok[e] the court’s authority ... to challenge the action sought to be adjudicated.” Valley Forge Christian College, v. Americans United for Separation of Church and State, supra, at 471-472. See, e. g., Flast v. Cohen, 392 U. S., at 101 (stake in “the dispute to be adjudicated in the lawsuit”); Eisenstadt v. Baird, 405 U. S. 438, 443 (1972) (plaintiff must have “sufficient interest in challenging the statute’s validity”).

The personal stake of a litigant depends, in turn, on whether he has alleged a legally redressable injury. In determining whether a plaintiff has a sufficient personal stake in the outcome of a controversy, this Court has asked whether he “personally has suffered some actual or threatened injury,” Gladstone, Realtors v. Village of Bellwood, 441 U. S. 91, 99 (1979) (emphasis added), whether the injury “fairly can be traced to the challenged action,” Simon v. Eastern Kentucky Welfare Rights Org., 426 U. S. 26, 41 (1976), and whether plaintiff’s injury “is likely to be redressed by a favorable decision.” Id., at 38. See also Duke Power Co. v. Carolina Environmental Study Group, Inc., 438 U. S. 59, 74 (1978); Warth v. Seldin, 422 U. S. 490, 508 (1975). These well-accepted criteria for determining whether a plaintiff has established the requisite personal stake do not fragment the standing inquiry into a series of discrete questions about the plaintiff’s stake in each of the particular types of relief sought. Quite the contrary, they ask simply whether the plaintiff has a sufficient stake in seeking a judicial resolution of the controversy.

Lyons has alleged past injury and a risk of future injury and has linked both to the city’s chokehold policy. Under established principles, the only additional question in determin*129ing standing under Art. Ill is whether the injuries he has alleged can be remedied or prevented by some form of judicial relief. Satisfaction of this requirement ensures that the lawsuit does not entail the issuance of an advisory opinion without the possibility of any judicial relief, and that the exercise of a court’s remedial powers will actually redress the alleged injury.20 Therefore Lyons needs to demonstrate only that, should he prevail on the merits, “the exercise of the Court’s remedial powers would redress the claimed injuries.” Duke Power Co., supra, at 74. See also Warth v. Seldin, supra, at 508; Simon, supra, at 38. Lyons has easily made this showing here, for monetary relief would plainly provide redress for his past injury, and prospective relief would reduce the likelihood of any future injury. Nothing more has ever been required to establish standing.

The Court’s decision turns these well-accepted principles on their heads by requiring a separate standing inquiry with *130respect to each request for relief. Until now, questions concerning remedy were relevant to the threshold issue of standing only in the limited sense that some relief must be possible. The approach adopted today drastically alters the inquiry into remedy that must be made to determine standing.

2

The Court’s fragmentation of the standing inquiry is also inconsistent with the way the federal courts have treated remedial issues since the merger of law and equity. The federal practice has been to reserve consideration of the appropriate relief until after a determination of the merits, not to foreclose certain forms of relief by a ruling on the pleadings. The prayer for relief is no part of the plaintiff’s cause of action. See 2A J. Moore & J. Lucas, Moore’s Federal Practice ¶8.18, p. 8-216, and n. 13 (1983) (Moore), and cases cited therein; C. Wright, A. Miller, & M. Kane, Federal Practice and Procedure § 2664 (1983) (Wright, Miller, & Kane). Rather, “[the usual rule is] that where legal rights have been invaded, and a federal statute provides for a general right to sue for such invasion, federal courts may use any available remedy to make good the wrong done.” Bell v. Hood, 327 U. S. 678, 684 (1946) (footnote omitted).

Rule 54(c) of the Federal Rules of Civil Procedure specifically provides that “every final judgment shall grant the relief to which the party in whose favor it is rendered is entitled, even if the party has not demanded such relief in his pleadings.” The question whether a plaintiff has stated a claim turns not on “whether [he] has asked for the proper remedy but whether he is entitled to any remedy.” (Emphasis added.) Wright, Miller, & Kane §2664. This is fully consistent with the approach taken in our standing cases. Supra, at 128-129 and this page, and n. 20.

The Court provides no justification for departing from the traditional treatment of remedial issues and demanding a separate threshold inquiry into each form of relief a plaintiff seeks. It is anomalous to require a plaintiff to demonstrate *131“standing” to seek each particular form of relief requested in the complaint when under Rule 54(c) the remedy to which a party may be entitled need not even be demanded in the complaint.21 See Holt Civic Club v. Tuscaloosa, 439 U. S. 60, 65-66 (1978); Albemarle Paper Co. v. Moody, 422 U. S. 405, 424 (1975). The traditional federal practice is a sound one. Even if it appears highly unlikely at the outset of a lawsuit that a plaintiff will establish that he is entitled to a particular remedy, there are dangers inherent in any doctrine that permits a court to foreclose any consideration of that remedy by ruling on the pleadings that the plaintiff lacks standing to seek it. A court has broad discretion to grant appropriate equitable relief to protect a party who has been injured by unlawful conduct, as well as members of the class, from future injury that may occur if the wrongdoer is permitted to continue his unlawful actions. Where, as here, a plaintiff alleges both past injury and a risk of future injury and presents a concededly substantial claim that a defendant is implementing an unlawful policy, it will rarely be easy to decide with any certainty at the outset of a lawsuit that no equitable relief would be appropriate under any conceivable set of facts that he might establish in support of his claim.

In sum, the Court’s approach to standing is wholly inconsistent with well-established standing principles and clashes with our longstanding conception of the remedial powers of a court and what is necessary to invoke the authority of a court to resolve a particular dispute.

I — I <1

Apart from the question of standing, the only remaining question presented in the petition for certiorari is whether *132the preliminary injunction issued by the District Court must be set aside because it “constituted a substantial interference in the operation of a municipal police department.” Pet. for Cert. i.22 In my view it does not.

In the portion of its brief concerning this second question, the city argues that the District Court ignored the principles of federalism set forth in Rizzo v. Goode, 423 U. S. 362 (1976). Brief for Petitioner 40-47. The city’s reliance on Rizzo is misplaced. That case involved an injunction which “significantly revis[ed] the internal procedures of the Philadelphia police department.” 423 U. S., at 379. The injunction required the police department to adopt “ ‘a comprehensive program for dealing adequately with civilian complaints’ ” to be formulated in accordance with extensive “guidelines” established by the District Court. Id., at 369, quoting Council of Organizations on Phila. Police A. & R. v. Rizzo, 357 F. Supp. 1289, 1321 (1973). Those guidelines specified detailed revisions of police manuals and rules of procedure, as well as the adoption of specific procedures for processing, screening, investigating, and adjudicating citizen complaints. In addition, the District Court supervised the implementation of the comprehensive program, issuing detailed orders concerning the posting and distribution of the revised police procedures and the drawing up of a “Citizen’s Complaint Report” in a format designated by the court. The District Court also reserved jurisdiction to review the progress of the police department. 423 U. S., at 365, n. 2. This Court concluded that the sweeping nature of the injunc-tive relief was inconsistent with “the principles of federalism.” Id., at 380.

*133The principles of federalism simply do not preclude the limited preliminary injunction issued in this case. Unlike the permanent injunction at issue in Rizzo, the preliminary injunction involved here entails no federal supervision of the LAPD’s activities. The preliminary injunction merely forbids the use of chokeholds absent the threat of deadly force, permitting their continued use where such a threat does exist. This limited ban takes the form of a preventive injunction, which has traditionally been regarded as the least intrusive form of equitable relief. Moreover, the city can remove the ban by obtaining approval of a training plan. Although the preliminary injunction also requires the city to provide records of the uses of chokeholds to respondent and to allow the court access to such records, this requirement is hardly onerous, since the LAPD already maintains records concerning the use of chokeholds.

A district court should be mindful that “federal-court intervention in the daily operation of a large city’s police department ... is undesirable and to be avoided if at all possible.” Rizzo, swpra, at 381 (Blackmun, J., dissenting).23 The modest interlocutory relief granted in this case differs markedly, however, from the intrusive injunction involved in Rizzo, and simply does not implicate the federalism concerns *134that arise when a federal court undertakes to “supervise the functioning of the police department.” 423 U. S., at 380.

V

Apparently because it is unwilling to rely solely on its unprecedented rule of standing, the Court goes on to conclude that, even if Lyons has standing, “[t]he equitable remedy is unavailable.” Ante, at 111. The Court’s reliance on this alternative ground is puzzling for two reasons.

If, as the Court says, Lyons lacks standing under Art. Ill, the federal courts have no power to decide his entitlement to equitable relief on the merits. Under the Court’s own view of Art. Ill, the Court’s discussion in Part V is purely an advisory opinion.

In addition, the question whether injunctive relief is available under equitable principles is simply not before us. We granted certiorari only to determine whether Lyons has standing and whether, if so, the preliminary injunction must be set aside because it constitutes an impermissible interference in the operation of a municipal police department. We did not grant certiorari to consider whether Lyons satisfies the traditional prerequisites for equitable relief. See n. 22, supra.

Even if the issue had been properly raised, I could not agree with the Court’s disposition of it. With the single exception of Rizzo v. Goode, supra,24 all of the cases relied on by the Court concerned injunctions against state criminal proceedings. The rule of Younger v. Harris, 401 U. S. 37 (1971), that such injunctions can be issued only in extraordinary circumstances in which the threat of injury is “great and immediate,” id., at 46, reflects the venerable rule that equity will not enjoin a criminal prosecution, the fact that constitu*135tional defenses can be raised in such a state prosecution, and an appreciation of the friction that injunctions against state judicial proceedings may produce. See ibid.; Steffel v. Thompson, 415 U. S. 452, 462 (1974); 28 U. S. C. §2283.

Our prior decisions have repeatedly emphasized that where an injunction is not directed against a state criminal or quasi-criminal proceeding, “the relevant principles of equity, comity, and federalism” that underlie the Younger doctrine “have little force.” Steffel v. Thompson, supra, at 462, citing Lake Carriers’ Assn. v. MacMullan, 406 U. S. 498, 509 (1972). Outside the special context in which the Younger doctrine applies, we have held that the appropriateness of in-junctive relief is governed by traditional equitable considerations. See Doran v. Salem Inn, Inc., 422 U. S. 922, 930 (1975). Whatever the precise scope of the Younger doctrine may be, the concerns of comity and federalism that counsel restraint when a federal court is asked to enjoin a state criminal proceeding simply do not apply to an injunction directed solely at a police department.

If the preliminary injunction granted by the District Court is analyzed under general equitable principles, rather than the more stringent standards of Younger v. Harris, it becomes apparent that there is no rule of law that precludes equitable relief and requires that the preliminary injunction be set aside. “In reviewing such interlocutory relief, this Court may only consider whether issuance of the injunction constituted an abuse of discretion.” Brown v. Chote, 411 U. S. 452, 457 (1973).

The District Court concluded, on the basis of the facts before it, that Lyons was choked without provocation pursuant to an unconstitutional city policy. Supra, at 119. Given the necessarily preliminary nature of its inquiry, there was no way for the District Court to know the precise contours of the city’s policy or to ascertain the risk that Lyons, who had alleged that the policy was being applied in a discriminatory manner, might again be subjected to a chokehold. But in view of the Court’s conclusion that the unprovoked choking of *136Lyons was pursuant to a city policy, Lyons has satisfied “the usual basis for injunctive relief, ‘that there exists some cognizable danger of recurrent violation.’” Rondeau v. Mosinee Paper Corp., 422 U. S. 49, 59 (1975), quoting United States v. W. T. Grant Co., 345 U. S. 629, 633 (1953). The risk of serious injuries and deaths to other citizens also supported the decision to grant a preliminary injunction. Courts of equity have much greater latitude in granting injunctive relief “in furtherance of the public interest than . . . when only private interests are involved.” Virginian R. Co. v. Railway Employee, 300 U. S. 515, 552 (1937). See Wright, Miller, & Kane § 2948; 7 Moore. ¶ 65.04[1]. In this case we know that the District Court would have been amply justified in considering the risk to the public, for after the preliminary injunction was stayed, five additional deaths occurred prior to the adoption of a moratorium. See n. 3, supra. Under these circumstances, I do not believe that the District Court abused its discretion.

Indeed, this Court has approved of a decision that directed issuance of a permanent injunction in a similar situation. See Lankford v. Gelston, 364 F. 2d 197 (CA4 1966), cited with approval in Allee v. Medrano, 416 U. S. 802, 816, n. 9 (1974). See n. 15, supra. In Lankford, citizens whose houses had been searched solely on the basis of uncorroborated, anonymous tips sought injunctive relief. The Fourth Circuit, sitting en banc, held that the plaintiffs were entitled to an injunction against enforcement of the police department policy authorizing such searches, even though there was no evidence that their homes would be searched in the future. Lyons is no less entitled to seek injunctive relief. To hold otherwise is to vitiate “one of the most valuable features of equity jurisdiction, to anticipate and prevent a threatened injury, where the damages would be insufficient or irreparable.” Vicksburg Waterworks Co. v. Vicksburg, 185 U. S. 65, 82 (1902).

*137Here it is unnecessary to consider the propriety of a permanent injunction. The District Court has simply sought to protect Lyons and other citizens of Los Angeles pending a disposition of the merits. It will be time enough to consider the propriety of a permanent injunction when and if the District Court grants such relief.

. VI

The Court’s decision removes an entire class of constitutional violations from the equitable powers of a federal court. It immunizes from prospective equitable relief any policy that authorizes persistent deprivations of constitutional rights as long as no individual can establish with substantial certainty that he will be injured, or injured again, in the future. The Chief Justice asked in Bivens v. Six Unknown Fed. Narcotics Agents, 403 U. S. 388, 419 (1971) (dissenting opinion), “what would be the judicial response to a police order authorizing ‘shoot to kill’ with respect to every fugitive”? His answer was that it would be “easy to predict our collective wrath and outrage.” Ibid. We now learn that wrath and outrage cannot be translated into an order to cease the unconstitutional practice, but only an award of damages to those who are victimized by the practice and live to sue and to the survivors of those who are not so fortunate. Under the view expressed by the majority today, , if the police adopt a policy of “shoot to kill,” or a policy of shooting 1 out of 10 suspects, the federal courts will be powerless to enjoin its continuation. Cf. Linda R. S. v. Richard D., 410 U. S., at 621 (White, J., dissenting). The federal judicial power is now limited to levying a toll for such a systematic constitutional violation.

*401Justice Alito

delivered the opinion of the Court.

Section 702 of the Foreign Intelligence Surveillance Act of 1978, 50 U. S. C. § 1881a (2006 ed., Supp. V), allows the Attorney General and the Director of National Intelligence to acquire foreign intelligence information by jointly authorizing the surveillance of individuals who are not “United States persons”1 and are reasonably believed to be located outside the United States. Before doing so, the Attorney General and the Director of National Intelligence normally must obtain the Foreign Intelligence Surveillance Court’s approval. Respondents are United States persons whose work, they allege, requires them to engage in sensitive international communications with individuals who they believe are likely targets of surveillance under § 1881a. Respondents seek a declaration that § 1881a is unconstitutional, as well as an injunction against § 1881a-authorized surveillance. The question before us is whether respondents have Article III standing to seek this prospective relief.

Respondents assert that they can establish injury in fact because there is an objectively reasonable likelihood that their communications will be acquired under § 1881a at some point in the future. But respondents’ theory of future injury is too speculative to satisfy the well-established requirement that threatened injury must be “certainly impending.” E. g., Whitmore v. Arkansas, 495 U. S. 149, 158 (1990). And even if respondents could demonstrate that the threatened injury is certainly impending, they still would not be able *402to establish that this injury is fairly traceable to § 1881a. As an alternative argument, respondents contend that they are suffering present injury because the risk of § 1881a-authorized surveillance already has forced them to take costly and burdensome measures to protect the confidentiality of their international communications. But respondents cannot manufacture standing by choosing to make expenditures based on hypothetical future harm that is not certainly impending. We therefore hold that respondents lack Article III standing.

I

A

In 1978, after years of debate, Congress enacted the Foreign Intelligence Surveillance Act (FISA) to authorize and regulate certain governmental electronic surveillance of communications for foreign intelligence purposes. See 92 Stat. 1783, 50 U. S. C. § 1801 et seq.; 1 D. Kris & J. Wilson, National Security Investigations & Prosecutions §§3.1, 3.7 (2d ed. 2012) (hereinafter Kris & Wilson). In enacting FISA, Congress legislated against the backdrop of our decision in United States v. United States Dist. Court for Eastern Dist. of Mich., 407 U. S. 297 (1972) (Keith), in which we explained that the standards and procedures that law enforcement officials must follow when conducting “surveillance of ‘ordinary crime’ ” might not be required in the context of surveillance conducted for domestic national-security purposes. Id., at 322-323. Although the Keith opinion expressly disclaimed any ruling “on the scope of the President’s surveillance power with respect to the activities of foreign powers,” id., at 308, it implicitly suggested that a special framework for foreign intelligence surveillance might be constitutionally permissible, see id., at 322-323.

In constructing such a framework for foreign intelligence surveillance, Congress created two specialized courts. In FISA, Congress authorized judges of the Foreign Intelli*403gence Surveillance Court (FISC) to approve electronic surveillance for foreign intelligence purposes if there is probable cause to believe that “the target of the electronic surveillance is a foreign power or an agent of a foreign power,” and that each of the specific “facilities or places at which the electronic surveillance is directed is being used, or is about to be used, by a foreign power or an agent of a foreign power.” § 105(a)(3), 92 Stat. 1790; see §§ 105(b)(1)(A), (b)(1)(B), ibid.; 1 Kris & Wilson §7:2, at 194-195; id., § 16:2, at 528-529. Additionally, Congress vested the Foreign Intelligence Surveillance Court of Review with jurisdiction to review any denials by the FISC of applications for electronic surveillance. § 103(b), 92 Stat. 1788; 1 Kris & Wilson §5:7, at 151-153.

In the wake of the September 11th attacks, President George W. Bush authorized the National Security Agency (NSA) to conduct warrantless wiretapping of telephone and e-mail communications where one party to the communication was located outside the United States and a participant in “the call was reasonably believed to be a member or agent of al Qaeda or an affiliated terrorist organization,” App. to Pet. for Cert. 403a. See id., at 263a-265a, 268a, 273a-279a, 292a-293a; American Civil Liberties Union v. NSA, 493 F. 3d 644, 648 (CA6 2007) (ACLU) (opinion of Batchelder, J.). In January 2007, the FISC issued orders authorizing the Government to target international communications into or out of the United States where there was probable cause to believe that one participant to the communication was a member or agent of al Qaeda or an associated terrorist organization. App. to Pet. for Cert. 3Í2a, 398a, 405a. These FISC orders subjected any electronic surveillance that was then occurring under the NSA’s program to the approval of the FISC. Id., at 405a; see id., at 312a, 404a. After a FISC Judge subsequently narrowed the FISC’s authorization of such surveillance, however, the Executive asked Congress to amend FISA so that it would provide the intelligence community with additional authority to meet the challenges *404of modern technology and international terrorism. Id., at 315a-318a, 331a-333a, 398a; see id., at 262a, 277a-279a, 287a.

When Congress enacted the FISA Amendments Act of 2008 (FISA Amendments Act), 122 Stat. 2436, it left much of FISA intact, but it “established a new and independent source of intelligence collection authority, beyond that granted in traditional FISA.” 1 Kris & Wilson §9:11, at 349-360. As relevant here, § 702 of FISA, 50 U. S. C. § 1881a (2006 ed., Supp. V), which was enacted as part of the FISA Amendments Act, supplements pre-existing FISA authority by creating a new framework under which the Government may seek the FISC’s authorization of certain foreign intelligence surveillance targeting the communications of non-U. S. persons located abroad. Unlike traditional FISA surveillance, § 1881a does not require the Government to demonstrate probable cause that the target of the electronic surveillance is a foreign power or agent of a foreign power. Compare §§ 1805(a)(2)(A), (a)(2)(B), with §§ 1881a(d)(l), (i)(3)(A); 638 F. 3d 118, 126 (CA2 2011); 1 Kris & Wilson § 16:16, at 584. And, unlike traditional FISA, § 1881a does not require the Government to specify the nature and location of each of the particular facilities or places at which the electronic surveillance will occur. Compare §§ 1805(a)(2)(B), (c)(1) (2006 ed. and Supp. V) with §§ 1881a(d)(l), (g)(4), (i)(3)(A); 638 F. 3d, at 125-126; 1 Kris & Wilson §16:16, at 585.2

The present case involves a constitutional challenge to § 1881a. Surveillance under § 1881a is subject to statutory conditions, judicial authorization, congressional supervision, and compliance with the Fourth Amendment. Section 1881a provides that, upon the issuance of an order from the FISC, “the Attorney General and the Director of National Intelligence may authorize jointly, for a period of up to 1 year ..., the targeting of persons reasonably believed to be located *405outside the United States to acquire foreign intelligence information.” § 1881a(a). Surveillance under § 1881a may not be intentionally targeted at any person known to be in the United States or any U. S. person reasonably believed to be located abroad. §§ 1881a(b)(l)-(3); see also §1801(i). Additionally, acquisitions under § 1881a must comport with the Fourth Amendment. § 1881a(b)(5). Moreover, surveillance under § 1881a is subject to congressional oversight and several types of Executive Branch review. See § § 1881a(f )(2), (0; Amnesty Int’l USA v. McConnell, 646 F. Supp. 2d 633, 640-641 (SDNY 2009).

Section 1881a mandates that the Government obtain the FISC’s approval of “targeting” procedures, “minimization” procedures, and a governmental certification regarding proposed surveillance. §§ 1881a(a), (c)(1), (i)(2), (i)(3). Among other things, the Government’s certification must attest that (1) procedures are in place “that have been approved, have been submitted for approval, or will be submitted with the certification for approval by the [FISC] that are reasonably designed” to ensure that an acquisition is “limited to targeting persons reasonably believed to be located outside” the United States; (2) minimization procedures adequately restrict the acquisition, retention, and dissemination of nonpublic information about unconsenting U. S. persons, as appropriate; (3) guidelines have been adopted to ensure compliance with targeting limits and the Fourth Amendment; and (4) the procedures and guidelines referred to above comport with the Fourth Amendment. § 1881a(g)(2); see § 1801(h).

The FISC’s role includes determining whether the Government’s certification contains the required elements. Additionally, the court assesses whether the targeting procedures are “reasonably designed” (1) to “ensure that an acquisition ... is limited to targeting persons reasonably believed to be located outside the United States” and (2) to “prevent the intentional acquisition of any communication as *406to which the sender and all intended recipients are known . . . to be located in the United States.” § 1881a(i)(2)(B). The court analyzes whether the minimization procedures “meet the definition of minimization procedures under section 1801(h) . . . , as appropriate.” § 1881a(i)(2)(C). The court also assesses whether the targeting and minimization procedures are consistent with the statute and the Fourth Amendment. See § 1881a(i)(3)(A).3

B

Respondents are attorneys and human rights, labor, legal, and media organizations whose work allegedly requires them to engage in sensitive and sometimes privileged telephone and e-mail communications with colleagues, clients, sources, and other individuals located abroad. Respondents believe that some of the people with whom they exchange foreign intelligence information are likely targets of surveillance under § 1881a. Specifically, respondents claim that they communicate by telephone and e-mail with people the Government “believes or believed to be associated with terrorist organizations,” “people located in geographic areas that are a special focus” of the Government’s counterterrorism or diplomatic efforts, and activists who oppose governments that are supported by the United States Government. App. to Pet. for Cert. 399a.

Respondents claim that § 1881a compromises their ability to locate witnesses, cultivate sources, obtain information, and communicate confidential information to their clients. Respondents also assert that they “have ceased engaging” in certain telephone and e-mail conversations. Id., at 400a. *407According to respondents, the threat of surveillance will compel them to travel abroad in order to have in-person conversations. In addition, respondents declare that they have undertaken “costly and burdensome measures” to protect the confidentiality of sensitive communications. Ibid.

C

On the day when the FISA Amendments Act was enacted, respondents filed this action seeking (1) a declaration that § 1881a, on its face, violates the Fourth Amendment, the First Amendment, Article III, and separation-of-powers principles and (2) a permanent injunction against the use of § 1881a. Respondents assért what they characterize as two separate theories of Article III standing. First, they claim that there is an objectively reasonable likelihood that their communications will be acquired under § 1881a at some point in the future, thus causing them injury. Second, respondents maintain that the risk of surveillance under § 1881a is so substantial that they have been forced to take costly and burdensome measures to protect the confidentiality of their international communications; in their view, the costs they have incurred constitute present injury that is fairly traceable to § 1881a.

After both parties moved for summary judgment, the District Court held that respondents do not have standing. 646 F. Supp. 2d, at 636. On appeal, however, a panel of the Second Circuit reversed. The panel agreed with respondents’ argument that they have standing due to the objectively reasonable likelihood that their communications will be intercepted at some time in the future. 638 F. 3d, at 133, 134, 139. In addition, the panel held that respondents have established that they are suffering “present injuries in fact— economic and professional harms—stemming from a reasonable fear of future harmful government conduct.” Id., at 138. The Second Circuit denied rehearing en banc by an equally divided vote. 667 F. 3d 163 (2011).

*408Because of the importance of the issue and the novel view of standing adopted by the Court of Appeals, we granted certiorari, 566 U. S. 1009 (2012), and we now

II

Article III of the Constitution limits federal courts’ jurisdiction to certain “Cases” and “Controversies.” As we have explained, “[n]o principle is more fundamental to the judiciary’s proper role in our system of government than the constitutional limitation of federal-court jurisdiction to actual cases or controversies.” DaimlerChrysler Corp. v. Cuno, 547 U. S. 332, 341 (2006) (internal quotation marks omitted); Raines v. Byrd, 521 U. S. 811, 818 (1997) (internal quotation marks omitted); see, e. g., Summers v. Earth Island Institute, 555 U. S. 488, 492-493 (2009). “One element of the case-or-eontroversy requirement” is that plaintiffs “must establish that they have standing to sue.” Raines, supra, at 818; see also Summers, supra, at 492-493; DaimlerChrysler Corp., supra, at 342; Lujan v. Defenders of Wildlife, 504 U. S. 555, 560 (1992).

The law of Article III standing, which is built on separation-of-powers principles, serves to prevent the judicial process from being used to usurp the powers of the political branches. Summers, supra, at 492-493; Daimler-Chrysler Corp., supra, at 341-342, 353; Raines, supra, at 818-820; Valley Forge Christian College v. Americans United for Separation of Church and State, Inc., 454 U. S. 464, 471-474 (1982); Schlesinger v. Reservists Comm. to Stop the War, 418 U. S. 208, 221-222 (1974). In keeping with the purpose of this doctrine, “our standing inquiry has been especially rigorous when reaching the merits of the dispute would force us to decide whether an action taken by one of the other two branches of the Federal Government was unconstitutional.” Raines, supra, at 819-820; see Valley Forge Christian College, supra, at 473-474; Schlesinger, supra, at 221-222. “Relaxation of standing re*409quirements is directly related to the expansion of judicial power,” United States v. Richardson, 418 U. S. 166, 188 (1974) (Powell, J., concurring); see also Summers, supra, at 492-493; Schlesinger, supra, at 222, and we have often found a lack of standing in cases in which the Judiciary has been requested to review actions of the political branches in the fields of intelligence gathering and foreign affairs, see, e. g., Richardson, supra, at 167-170 (plaintiff lacked standing to challenge the constitutionality of a statute permitting the Central Intelligence Agency to account for its expenditures solely on the certificate of the CIA Director); Schlesinger, supra, at 209-211 (plaintiffs lacked standing to challenge the Armed Forces Reserve membership of Members of Congress); Laird v. Tatum, 408 U. S. 1, 11-16 (1972) (plaintiffs lacked standing to challenge an Army intelligence-gathering program).

To establish Article III standing, an injury must be “concrete, particularized, and actual or imminent; fairly traceable to the challenged action; and redressable by a favorable ruling.” Monsanto Co. v. Geertson Seed Farms, 561 U. S. 139, 149 (2010); see also Summers, supra, at 493; Defenders of Wildlife, 504 U. S., at 560-561. “Although imminence is coneededly a somewhat elastic concept, it cannot be stretched beyond its purpose, which is to ensure that the alleged injury is not too speculative for Article III purposes—that the injury is certainly impending.” Id., at 565, n. 2 (internal quotation marks omitted). Thus, we have repeatedly reiterated that “threatened injury must be certainly impending to constitute injury in fact,” and that “[allegations of possible future injury” are not sufficient. Whitmore, 495 U. S., at 158 (emphasis added; internal quotation marks omitted); see also Defenders of Wildlife, supra, at 565, n. 2, 567, n. 3; see Daimler Chrysler Corp., supra, at 345; Friends of the Earth, Inc. v. Laidlaw Environmental Services (TOC), Inc., 528 U. S. 167, 190 (2000); Babbitt v. Farm Workers, 442 U. S. 289, 298 (1979).

*410i—I 1—I H-4

A

Respondents assert that they can establish injury in fact that is fairly traceable to § 1881a because there is an objectively reasonable likelihood that their communications with their foreign contacts will be intercepted under § 1881a at some point in the future. This argument fails. As an initial matter, the Second Circuit’s “objectively reasonable likelihood” standard is inconsistent with our requirement that “threatened injury must be certainly impending to constitute injury in fact.” Whitmore, supra, at 158 (internal quotation marks omitted); see also Daimler Chrysler Corp., supra, at 345; Laidlaw, supra, at 190; Defenders of Wildlife, supra, at 565, n. 2; Babbitt, supra, at 298. Furthermore, respondents’ argument rests on their highly speculative fear that: (1) the Government will decide to target the communications of non-U. S. persons with whom they communicate; (2) in doing so, the Government will choose to invoke its authority under § 1881a rather than utilizing another method of surveillance; (3) the Article III judges who serve on the FISC will conclude that the Government’s proposed surveillance procedures satisfy § 188 la’s many safeguards and are consistent with the Fourth Amendment; (4) the Government will succeed in intercepting the communications of respondents’ contacts; and (5) respondents will be parties to the particular communications that the Government intercepts. As discussed below, respondents’ theory of standing, which relies on a highly attenuated chain of possibilities, does not satisfy the requirement that threatened injury must be certainly impending. See Summers, supra, at 496 (rejecting a standing theory premised on a speculative chain of possibilities); Whitmore, supra, at 157-160 (same). Moreover, even if respondents could demonstrate injury in fact, the second link in the above-described chain of contingencies—which amounts to mere speculation about whether surveillance *411would be under § 1881a or some other authority—shows that respondents cannot satisfy the requirement that any injury in fact must be fairly traceable to § 1881a.

First, it is speculative whether the Government will imminently target communications to which respondents are parties. Section 1881a expressly provides that respondents, who are U. S. persons, cannot be targeted for surveillance under § 1881a. See §§ 1881a(b)(1)-(3); 667 F. 3d, at 173 (Raggi, J., dissenting from denial of rehearing en banc). Accordingly, it is no surprise that respondents fail to offer any evidence that their communications have been monitored under § 1881a, a failure that substantially undermines their standing theory. See ACLU, 493 F. 3d, at 655-656, 673-674 (opinion of Batchelder, J.) (concluding that plaintiffs who lacked evidence that their communications had been intercepted did not have standing to challenge alleged NS A surveillance). Indeed, respondents do not even allege that the Government has sought the FISC’s approval for surveillance of their communications. Accordingly, respondents’ theory necessarily rests on their assertion that the Government will target other individuals—namely, their foreign contacts.

Yet respondents have no actual knowledge of the Government’s § 1881a targeting practices. Instead, respondents merely speculate and make assumptions about whether their communications with their foreign contacts will be acquired under § 1881a. See 667 F. 3d, at 185-187 (opinion of Raggi, J.). For example, journalist Christopher Hedges states: “I have no choice but to assume that any of my international communications may be subject to government surveillance, and I have to make decisions ... in light of that assumption.” App. to Pet. for Cert. 366a (emphasis added and deleted). Similarly, attorney Scott McKay asserts that, “[bjecause of the [FISA Amendments Act], we now have to assume that every one of our international communications may be monitored by the government.” Id., at 375a (emphasis added); see also id., at 337a, 343a-344a, 350a, 356a. “The party in-*412yoking federal jurisdiction bears the burden of establishing” standing—and, at the summary judgment stage, such a party “can no longer rest on . . . ‘mere allegations,’ but must ‘set forth’ by affidavit or other evidence ‘specific facts.’” Defenders of Wildlife, 504 U. S., at 561. Respondents, however, have set forth no specific facts demonstrating that the communications of their foreign contacts will be targeted. Moreover, because § 1881a at most authorizes—but does not mandate or direct—the surveillance that respondents fear, respondents’ allegations are necessarily conjectural. See United Presbyterian Church in U. S. A. v. Reagan, 738 F. 2d 1375, 1380 (CADC 1984) (Scalia, J.); 667 F. 3d, at 187 (opinion of Raggi, J.). Simply put, respondents can only speculate as to how the Attorney General and the Director of National Intelligence will exercise their discretion in determining which communications to target.4

Second, even if respondents could demonstrate that the targeting of their foreign contacts is imminent, respondents can only speculate as to whether the Government will seek to use § 1881a-authorized surveillance (rather than other methods) to do so. The Government has numerous other *413methods of conducting surveillance, none of which is challenged here. Even after the enactment of the FISA Amendments Act, for example, the Government may still conduct electronic surveillance of persons abroad under the older provisions óf FISA so long as it satisfies the applicable requirements, including a demonstration of probable cause to believe that the person is a foreign power or agent of a foreign power. See § 1805. The Government may also obtain information from the intelligence services of foreign nations. Brief for Petitioners 33. And, although we do not reach the question, the Government contends that it can conduct FISA-exempt human and technical surveillance programs that are governed by Executive Order 12333. See Exec. Order No. 12333, §§ 1.4, 2.1-2.5, 3 CFR 202, 210-212 (1981), reprinted as amended, note following 50 U. S. C. §401, pp. 543, 547-548. Even if respondents could demonstrate that their foreign contacts will imminently be targeted— indeed, even if they could show that interception of their own communications will imminently occur—they would still need to show that their injury is fairly traceable to § 1881a. But, because respondents can only speculate as to whether any (asserted) interception would be under § 1881a or some other authority, they cannot satisfy the “fairly traceable” requirement.

Third, even if respondents could show that the Government will seek the FISC’s authorization to acquire the communications of respondents’ foreign contacts under § 1881a, respondents can only speculate as to whether that court will authorize such surveillance. In the past, we have been reluctant to endorse standing theories that require guesswork as to how independent decisionmakers will exercise their judgment. In Whitmore, for example, the plaintiff’s theory of standing hinged largely on the probability that he would obtain federal habeas relief and be convicted upon retrial. In holding that the plaintiff lacked standing, we explained that “[i]t is just not possible for a litigant to prove in advance *414that the judicial system will lead to any particular result in his case.” 495 U. S., at 159-160; see Defenders of Wildlife, supra, at 562.

We decline to abandon our usual reluctance to endorse standing theories that rest on speculation about the decisions of independent actors. Section 1881a mandates that the Government must obtain the FISC’s approval of targeting procedures, minimization procedures, and a governmental certification regarding proposed surveillance. §§ 1881a(a), (e)(1), (i)(2), (i)(3). The court must, for example, determine whether the Government’s procedures are “reasonably designed ... to minimize the acquisition and retention, and prohibit the dissemination, of nonpublicly available information concerning unconsenting United States persons.” § 1801(h); see §§ 1881a(i)(2), (i)(3)(A). And, critically, the court must also assess whether the Government’s targeting and minimization procedures comport with the Fourth Amendment. § 1881a(i)(3)(A).

Fourth, even if the Government were to obtain the FISC’s approval to target respondents’ foreign contacts under § 1881a, it is unclear whether the Government would succeed in acquiring the communications of respondents’ foreign contacts. And fifth, even if the Government were to conduct surveillance of respondents’ foreign contacts, respondents can only speculate as to whether their own communications with their foreign contacts would be incidentally acquired.

In sum, respondents’ speculative chain of possibilities does not establish that injury based on potential future surveillance is certainly impending or is fairly traceable to § 1881a.5

*415B

Respondents’ alternative argument—namely, that they can establish standing based on the measures that they have undertaken to avoid § 1881a-authorized surveillance—fares no better. Respondents assert that they are suffering ongoing injuries that are fairly traceable to § 1881a because the risk of surveillance under § 1881a requires them to take' costly and burdensome measures to protect the confidentiality of their communications. Respondents claim, for instance, that the threat of surveillance sometimes compels them to avoid certain e-mail and phone conversations, to “tal[k] in generalities rather than specifics,” or to travel so that they can have in-person conversations. Tr. of Oral Arg. 38; App. to Pet. for Cert. 338a, 345a, 367a, 400a.6 The Second Circuit panel concluded that, because respondents are already suffering such ongoing injuries, the likelihood of interception under § 1881a is relevant only to the question whether respondents’ ongoing injuries are “fairly traceable” to § 1881a. See 638 F. 3d, at 133-134; 667 F. 3d, at 180 (opinion of Raggi, J.). Analyzing the “fairly traceable” element of standing under a relaxed reasonableness standard, see 638 F. 3d, at 133-134, the Second Circuit then held that “plaintiffs *416have established that they suffered present injuries in fact— economic and professional harms—stemming from a reasonable fear oí future harmful government conduct,” id., at 138.

The Second Circuit’s analysis improperly allowed respondents to establish standing by asserting that they suffer present costs and burdens that are based on a fear of surveillance, so long as that fear is not “fanciful, paranoid, or otherwise unreasonable.” See id., at 134. This improperly waters down the fundamental requirements of Article III. Respondents’ contention that they have standing because they incurred certain costs as a reasonable reaction to a risk of harm is unavailing—because the harm respondents seek to avoid is not certainly impending. In other words, respondents cannot manufacture standing merely by inflicting harm on themselves based on their fears of hypothetical future harm that is not certainly impending. See Pennsylvania v. New Jersey, 426 U. S. 660, 664 (1976) (per curiam); National Family Planning & Reproductive Health Assn., Inc. v. Gonzales, 468 F. 3d 826, 831 (CADC 2006). Any ongoing injuries that respondents are suffering are not fairly traceable to § 1881a.

If the law were otherwise, an enterprising plaintiff would be able to secure a lower standard for Article III standing simply by making an expenditure based on a nonparanoid fear. As Judge Raggi accurately noted, under the Second Circuit panel’s reasoning, respondents could, “for the price of a plane ticket, . . . transform their standing burden from one requiring a showing of actual or imminent. . . interception to one requiring a showing that their subjective fear of such interception is not fanciful, irrational, or clearly unreasonable.” 667 F. 3d, at 180 (internal quotation marks omitted). Thus, allowing respondents to bring this action based on costs they incurred in response to a speculative threat would be tantamount to accepting a repackaged version of respondents’ first failed theory of standing. See ACLU, 493 F. 3d, at 656-657 (opinion of Batchelder, J.).

*417Another reason that respondents’ present injuries are not fairly traceable to § 1881a is that even before § 1881a was enacted, they had a similar incentive to engage in many of the countermeasures that they are now taking. See id., at 668-670. For instance, respondent Scott McKay’s declaration describes—and the dissent heavily relies on—McKay’s “knowledge” that thousands of communications involving one of his clients were monitored in the past. App. to Pet. for Cert. 370a; post, at 426-426, 429. But this surveillance was conducted pursuant to FISA authority that predated § 1881a. See Brief for Petitioners 32, n. 11; Al-Kidd v. Gonzales, No. 06-cv-93, 2008 WL 5123009 (D Idaho, Dec. 4, 2008). Thus, because the Government was allegedly conducting surveillance of McKay’s client before Congress enacted § 1881a, it is difficult to see how the safeguards that McKay now claims to have implemented can be traced to § 1881a.

Because respondents do not face a threat of certainly impending interception under § 1881a, the costs that they have incurred to avoid surveillance are simply the product of their fear of surveillance,7 and our decision in Laird makes it clear that such a fear is insufficient to create standing. See 408 U. S., at 10-15. The plaintiffs in Laird argued that their exercise of First Amendment rights was being “chilled by the mere existence, without more, of [the Army’s] investigative and data-gathering activity.” Id., at 10. While acknowledging that prior cases had held that constitutional violations may arise from the chilling effect of “regulations that fall short of a direct prohibition against the exercise of *418First Amendment rights,” the Court declared that none of those cases involved a “chilling effect arising] merely from the individual’s knowledge that a governmental agency was engaged in certain activities or from the individual’s concomitant fear that, armed with the fruits of those activities, the agency might in the future take some other and additional action detrimental to that individual.” Id., at 11. Because “[allegations of a subjective ‘chill’ are not an adequate substitute for a claim of specific present objective harm or a threat of specific future harm,” id., at 13-14, the plaintiffs in Laird—and respondents here—lack standing. See ibid.) ACLU, supra, at 661-662 (opinion of Batchelder, J.) (holding that plaintiffs lacked standing because they “allege[d] only a subjective apprehension” of alleged NS A surveillance and “a personal (self-imposed) unwillingness to communicate”); United Presbyterian Church, 738 F. 2d, at 1378 (holding that plaintiffs lacked standing to challenge the legality of ah Executive Order relating to surveillance because “the ‘chilling effect’ which is produced by their fear of being subjected to illegal surveillance and which deters them from conducting constitutionally protected activities, is foreclosed as a basis for standing” by Laird).

For the reasons discussed above, respondents’ self-inflicted injuries are not fairly traceable to the Government’s purported activities under § 1881a, and their subjective fear of surveillance does not give rise to standing.

IV

A

Respondents incorrectly maintain that “[t]he kinds of injuries incurred here—injuries incurred because of [respondents’] reasonable efforts to avoid greater injuries that are otherwise likely to flow from the conduct they challenge— are the same kinds of injuries that this Court held to support. standing in cases such as” Laidlaw, Meese v. Keene, 481 U. S. *419465 (1987), and Monsanto. Brief for Respondents 24. As an initial matter, none of these cases holds or even suggests that plaintiffs can establish standing simply by claiming that they .experienced a “chilling effect” that resulted from a governmental policy that does not regulate, constrain, or compel any action on their part. Moreover, each of these cases was very different from the present case.

In Laidlaw, plaintiffs’ standing was based on “the proposition that a company’s continuous and pervasive illegal discharges of pollutants into a river would cause nearby residents to curtail their recreational use of that waterway and would subject them to other economic and aesthetic harms.” 528 U. S., at 184. Because the unlawful discharges of pollutants were “concededly ongoing,” the only issue was whether “nearby residents”—who were members of the organizational plaintiffs—acted reasonably in refraining from using the polluted area. Id., at 183-184. Laidlaw is therefore quite unlike the present case, in which it is not “concede[d]” that respondents would be subject to unlawful surveillance but for their decision to take preventive measures. See ACLU, 493 F. 3d, at 686 (opinion of Batchelder, J.) (distinguishing Laidlaw on this ground); id., at 689-690 (Gibbons, J., concurring) (same); 667 F. 3d, at 182-183 (opinion of Raggi, J.) (same). Laidlaw would resemble this case only if (1) it were undisputed that the Government was using § 1881a-authorized surveillance to acquire respondents’ communications and (2) the sole dispute concerned the reasonableness of respondents’ preventive measures.

In Keene, the plaintiff challenged the constitutionality of the Government’s decision to label three films as “political propaganda.” 481 U. S., at 467. The Court held that the plaintiff, who was an attorney and a state legislator, had standing because he demonstrated, through “detailed affidavits,” that he “could not exhibit the films without incurring a risk of injury to his reputation and of an impairment of his *420political career.” Id., at 467, 473-475. Unlike the present case, Keene involved “more than a ‘subjective chill’” based on speculation about potential governmental action; the plaintiff in that case was unquestionably regulated by the relevant statute, and the films that he wished to exhibit had already been labeled as “political propaganda.” See ibid.; ACLU, 493 F. 3d, at 663-664 (opinion of Batchelder, J.); id., at 691 (Gibbons, J., concurring).

Monsanto, on which respondents also rely, is likewise in-apposite. In Monsanto, conventional alfalfa farmers had standing to seek injunctive relief because the agency’s decision to deregulate a variety of genetically engineered alfalfa gave rise to a “significant risk of gene flow to non-genetically-engineered varieties of alfalfa.” 561 U. S., at 155. The standing analysis in that case hinged on evidence that genetically engineered alfalfa “ ‘seed fields [we]re currently being planted in all the major alfalfa seed production areas’ ”; the bees that pollinate alfalfa “ ‘have a range of at least two to ten miles’ and the alfalfa seed farms were concentrated in an area well within the bees’ pollination range. Id., at 154, and n. 3. Unlike the conventional alfalfa farmers in Monsanto, however, respondents in the present case present no concrete evidence to substantiate their fears, but instead rest on mere conjecture about possible governmental actions.

B

Respondents also suggest that they should be held to have standing because otherwise the constitutionality of § 1881a could not be challenged. It would be wrong, they maintain, to “insulate the government’s surveillance activities from meaningful judicial review.” Brief for Respondents 60. Respondents’ suggestion is both legally and factually incorrect. First, “ ‘[t]he assumption that if respondents have no standing to sue, no one would have standing, is not a reason to find standing.’” Valley Forge Christian College, 454 U. S., at 489; Schlesinger, 418 U. S., at 227; see also Richard*421son, 418 U. S., at 179; Raines, 521 U. S., at 835 (Souter, J., joined by Ginsburg, J., concurring in judgment).

Second, our holding today by no means insulates § 1881a from judicial review. As described above, Congress created a comprehensive scheme in which the FISC evaluates the Government’s certifications, targeting procedures, and minimization procedures—including assessing whether the targeting and minimization procedures comport with the Fourth Amendment. §§1881a(a), (c)(1), (i)(2), (i)(3). Any dissatisfaction that respondents may have about the FISC’s rulings—or the congressional delineation of that court’s role—is irrelevant to our standing analysis.

Additionally, if the Government intends to use or disclose information obtained or derived from a § 1881a acquisition in judicial or administrative proceedings, it must provide advance notice of its intent, and the affected person may challenge the lawfulness of the acquisition. §§ 1806(c), (e), 1881e(a) (2006 ed. and Supp. V).8 Thus, if the Government were to prosecute one of respondent-attorney’s foreign clients using § 188 la-authorized surveillance, the Government would be required to make a disclosure. Although the foreign client might not have a viable Fourth Amendment claim, see, e. g., United States v. Verdugo-Urquidez, 494 U. S. 259, 261 (1990), it is possible that the monitoring of the target’s conversations with his or her attorney would provide grounds for a claim of standing on the part of the attorney. Such an attorney would certainly have a stronger eviden-tiary basis for establishing standing than do respondents in the present case. In such a situation., unlike in the pres*422ent case, it would at least be clear that the Government had acquired the foreign client’s communications using § 188 la-authorized surveillance.

Finally, any electronic communications service provider that the Government directs to assist in § 1881a surveillance may challenge the lawfulness of that directive before the FISC. §§ 1881a(h)(4), (h)(6). Indeed, at the behest of a service provider, the Foreign Intelligence Surveillance Court of Review previously analyzed the constitutionality of electronic surveillance directives issued pursuant to a now-expired set of FISA amendments. See In re Directives Pursuant to Section 105B of Foreign Intelligence Surveillance Act, 551 F. 3d 1004, 1006-1016 (2008) (holding that the provider had standing and that the directives were constitutional).

* * *

We hold that respondents lack Article III standing because they cannot demonstrate that the future injury they purportedly fear is certainly impending and because they cannot manufacture standing by incurring costs in anticipation of nonimminent harm. We therefore reverse the judgment of the Second Circuit and remand the case for further proceedings consistent with this opinion.

It is so ordered.

Justice Breyer,

with whom Justice Ginsburg, Justice Sotomayor, and Justice Kagan join, dissenting.

The plaintiffs’ standing depends upon the likelihood that the Government, acting under the authority of 50 U. S. C. § 1881a (2006 ed., Supp. V), will harm them by intercepting at least some of their private, foreign, telephone, or e-mail conversations. In my view, this harm is not “speculative.” Indeed it is as likely to take place as are most future events that commonsense inference and ordinary knowledge of human nature tell us will happen. This Court has often found the occurrence of similar future events sufficiently cer*423tain to support standing. I dissent from the Court’s contrary conclusion.

I

Article III specifies that the “judicial Power” of the United States extends only to actual “Cases” and “Controversies.” § 2. It thereby helps to ensure that the legal questions presented to the federal courts will not take the form of abstract intellectual problems resolved in the “rarified atmosphere of a debating society” but instead those questions will be presented “in a concrete factual context conducive to a realistic appreciation of the consequences of judicial action.” Valley Forge Christian College v. Americans United for Separation of Church and State, Inc., 454 U. S. 464, 472 (1982) (purpose of Article III); Lujan v. Defenders of Wildlife, 504 U. S. 555, 560 (1992) (similar); Babbitt v. Farm Workers, 442 U. S. 289, 297 (1979) (similar).

The Court has recognized that the precise boundaries of the “case or controversy” requirement are matters of “degree ... not discernible by any precise test.” Ibid. At the same time, the Court has developed a subsidiary set of legal rules that help to determine when the Constitution’s requirement is met. See Lujan, 504 U. S., at 560-561; id., at 583 (Stevens, J., concurring in judgment). Thus, a plaintiff must have “standing” to bring a legal claim. And a plaintiff has that standing, the Court has said, only if the action or omission that the plaintiff challenges has caused, or will cause, the plaintiff to suffer an injury that is “concrete and particularized,” “actual or imminent,” and “redressfcble] by a favorable decision.” Id., at 560-561 (internal quotation marks omitted).

No one here denies that the Government’s interception of a private telephone or e-mail conversation amounts to an injury that is “concrete and particularized.” Moreover, the plaintiffs, respondents here, seek as relief a judgment declaring unconstitutional (and enjoining enforcement of) a statutory provision authorizing those interceptions; and, such a *424judgment would redress the injury by preventing it. Thus, the basic question is whether the injury, i. e., the interception, is “actual or imminent.”

II

A

Since the plaintiffs fear interceptions of a kind authorized by § 1881a, it is important to understand just what kind of surveillance that section authorizes. Congress enacted § 1881a in 2008, as an amendment to the pre-existing Foreign Intelligence Surveillance Act of 1978, 50 U. S. C. § 1801 et seq. Before the amendment, the Act authorized the Government (acting within the United States) to monitor private electronic communications between the United States and a foreign country if (1) the Government’s purpose was, in significant part, to obtain foreign intelligence information (which includes information concerning a “foreign power” or “territory” related to our “national defense” or “security” or the “conduct of . . . foreign affairs”), (2) the Government’s surveillance target was “a foreign power or an agent of a foreign power,” and (3) the Government used surveillance procedures designed to “minimize the acquisition and retention, and prohibit the dissemination, of” any private information acquired about Americans. §§ 1801(e), (h), 1804(a).

In addition, the Government had to obtain the approval of the Foreign Intelligence Surveillance Court. To do so, it had to submit an application describing (1) each “specific target,” (2) the “nature of the information sought,” and (3) the “type of communications or activities to be subjected to the surveillance.” § 1804(a). It had to certify that, in significant part, it sought to obtain foreign intelligence information. Ibid. It had to demonstrate probable cause to believe that each specific target was “a foreign power or an agent of a foreign power.” §§ 1804(a), 1805(a). It also had to describe instance-specific procedures to be used to minimize intrusions upon Americans’ privacy (compliance *425with which the court subsequently could assess). §§ 1804(a), 1805(d)(3).

The addition of § 1881a in 2008 changed this prior law in three important ways. First, it eliminated the requirement that the Government describe to the court each specific target and identify each facility at which its surveillance would be directed, thus permitting surveillance on a programmatic, not necessarily individualized, basis. § 1881a(g). Second, it eliminated the requirement that a target be a “foreign power or an agent of a foreign power.” Ibid. Third, it diminished the court’s authority to insist upon, and eliminated its authority to supervise, instance-specific privacy-intrusion minimization procedures (though the Government still must use court-approved general minimization procedures). § 1881a(e). Thus, using the authority of § 1881a, the Government can obtain court approval for its surveillance of electronic communications between places within the United States and targets in foreign territories by showing the court (1) that “a significant purpose of the acquisition is to obtain foreign intelligence information,” and (2) that it will use general targeting and privacy-intrusion minimization procedures of a kind that the court had previously approved. § 1881a(g).

B

It is similarly important to understand the kinds of communications in which the plaintiffs say they engage and which they believe the Government will intercept. Plaintiff Scott McKay, for example, says in an affidavit (1) that he is a lawyer; (2) that he represented “Mr. Sami Omar Al-Hussayen, who was acquitted in June 2004 on terrorism charges”; (3) that he continues to represent “Mr. Al-Hussayen, who, in addition to facing criminal charges after September 11, was named as a defendant in several civil cases”; (4) that he represents Khalid Sheik Mohammed, a detainee, “before the Military Commissions at Guantánamo Bay, Cuba”; (5) that in representing these clients he “commu-*426nicatefe] by telephone and email with people outside the United States, including Mr. Al-Hussayen himself,” “experts, investigators, attorneys, family members ... and others who are located abroad”; and (6) that prior to 2008 “the U. S. government had intercepted some 10,000 telephone calls and 20,000 email communications involving [his client] Al-Hussayen.” App. to Pet. for Cert. 369a-371a.

Another plaintiff, Sylvia Royce, says in her affidavit (1) that she is an attorney; (2) that she “represent^] Moham-medou Ould Salahi, a prisoner who has been held at Guantá-namo Bay as an enemy combatant”; (3) that, “[i]n connection with [her] representation of Mr. Salahi, [she] received] calls from time to time from Mr. Salahi’s brother,... a university student in Germany”; and (4) that she has been told that the Government has threatened Salahi “that his family members would be arrested and mistreated if he did not cooperate.” Id., at 349a-351a.

The plaintiffs have noted that McKay no longer represents Mohammed and Royce no longer represents Ould Salahi. Brief for Respondents 15, n. 11. But these changes are irrelevant, for we assess standing as of the time a suit is filed, see Davis v. Federal Election Comm’n, 554 U. S. 724, 734 (2008), and in any event McKay himself continues to represent Al Hussayen, his partner now represents Mohammed, and Royce continues to represent individuals held in the custody of the U. S. military overseas.

A third plaintiff, Joanne Mariner, says in her affidavit (1) that she is a human rights researcher; (2) that “some of the work [she] do[es] involves trying to track down people who were rendered by the CIA to countries in which they were tortured”; (3) that many of those people “the CIA has. said are (or were) associated with terrorist organizations”; and (4) that, to do this research, she “communicate[s] by telephone and e-mail with . . . former detainees, lawyers for detainees, relatives of detainees, political activists, journalists, and fixers” “all over the world, including in Jordan, Egypt, Paki*427stan, Afghanistan, [and] the Gaza Strip.” App. to Pet. for Cert. 343a-344a.

Other plaintiffs, including lawyers, journalists, and human rights researchers, say in affidavits (1) that they have jobs that require them to gather information from foreigners located abroad; (2) that they regularly communicate electronically (e. g., by telephone or e-mail) with foreigners located abroad; and (3) that in these communications they exchange “foreign intelligence information” as the Act defines it. Id., at 334a-375a.

HH H-I

Several considerations, based upon the record along with commonsense inferences, convince me that there is a very high likelihood that the Government, acting under the authority of § 1881a, will intercept at least some of the communications just described. First, the plaintiffs have engaged, and continue to engage, in electronic communications of a kind that the 2008 amendment, but not the prior Act, authorizes the Government to intercept. These communications include discussions with family members of those detained at Guantanamo, friends and acquaintances of those persons, and investigators, experts, and others with knowledge of circumstances related to terrorist activities. These persons are foreigners located outside the United States. They are not “foreign power[s]” or “agent[s] of . . . foreign powerfe].” And the plaintiffs state that they exchange with these persons “foreign intelligence information,” defined to include information that “relates to” “international terrorism” and “the national defense or the security of the United States.” See 50 U. S. C. § 1801 (2006 ed. and Supp. V); see, e. g., App. to Pet. for Cert. 342a, 366a, 373a-374a.

Second, the plaintiffs have a strong motive to engage in, and the Government has a strong motive to listen to, conversations of the kind described. A lawyer representing a client normally seeks to learn the circumstances surrounding the crime (or the civil wrong) of which the client is accused. *428A fair reading of the affidavit of Scott McKay, for example, taken together with elementary considerations of a lawyer’s obligation to his client, indicates that McKay will engage in conversations that concern what suspected foreign terrorists, such as his client, have done; in conversations that concern his clients’ families, colleagues, and contacts; in conversations that concern what those persons (or those connected to them) have said and done, at least in relation to terrorist activities; in conversations that concern the political, social, and commercial environments in which the suspected terrorists have lived and worked; and so forth. See, e. g., id., at 373a~374a. Journalists and human rights workers have strong similar motives to conduct conversations of this kind. See, e. g., id., at 342a (declaration of Joanne Mariner, stating that “some of the information [she] exchangefs] by telephone and e-mail relates to terrorism and counterterrorism, and much of the information relates to the foreign affairs of the United States”).

At the same time, the Government has a strong motive to conduct surveillance of conversations that contain material of this kind. The Government, after all, seeks to learn as much as it can reasonably learn about suspected terrorists (such as those detained at Guantanamo), as well as about their contacts and activities, along with those of friends and family members. See Executive Office of the President, Office of Management and Budget, Statement of Administration Policy on S. 2248, p. 4 (Dec. 17, 2007) (“Part of the value of the [new authority] is to enable the Intelligence Community to collect expeditiously the communications of terrorists in foreign countries who may contact an associate in the United States”). And the Government is motivated to do so, not simply by the desire to help convict those whom the Government believes guilty, but also by the critical, overriding need to protect America from terrorism. See id., at 1 (“Protection of the American people and American interests at home and abroad requires access to timely, accurate, and *429insightful intelligence on the capabilities, intentions, and activities of . . . terrorists”).

Third, the Government’s past behavior shows that it has sought, and hence will in all likelihood continue to seek, information about alleged terrorists and detainees through means that include surveillance of electronic communications. As just pointed out, plaintiff Scott McKay states that the Government (under the authority of the pre-2008 law) “intercepted some 10,000 telephone calls and 20,000 email communications involving [his client] Mr. Al-Hussayen.” App. to Pet. for Cert. 370a.

Fourth, the Government has the capacity to conduct electronic surveillance of the kind at issue. To some degree this capacity rests upon technology available to the Government. See 1 D. Kris & J. Wilson, National Security Investigations & Prosecutions § 16:6, p. 562 (2d ed. 2012) (“NSA’s technological abilities are legendary”); id., § 16:12, at 572-577 (describing the National Security Agency’s capacity to monitor “very broad facilities” such as international switches). See, e. g., Lichtblau & Risen, Spy Agency Mined Vast Data Trove, Officials Report, N. Y. Times, Dec. 24,' 2005, p. A1 (describing capacity to trace and to analyze large volumes of communications into and out of the United States); Lichtblau & Shane, Bush Is Pressed Over New Report on Surveillance, N. Y. Times, May 12, 2006, p. A1 (reporting capacity to obtain access to records of many, if not most, telephone calls made in the United States); Priest & Arkin, A Hidden World, Growing Beyond Control, Washington Post, July 19, 2010, p. A1 (reporting that every day, collection systems at the National Security Agency intercept and store 1.7 billion e-mails, telephone calls, and other types of communications). Cf. Statement of Administration Policy on S. 2248, supra, at 3 (rejecting a provision of the Senate bill that would require intelligence analysts to count “the number of persons located in the United States whose communications were reviewed” as “impossible to implement” (internal quotation marks *430omitted)). This capacity also includes the Government’s authority to obtain the kind of information here at issue from private carriers such as AT&T and Verizon. See 50 U. S. C. § 1881a(h). We are farther told by amici that the Government is expanding that capacity. See Brief for Electronic Privacy Information Center et al. 22-23 (National Security Agency will be able to conduct surveillance of most electronic communications between domestic and foreign points).

Of course, to exercise this capacity the Government must have intelligence court authorization. But the Government rarely files requests that fail to meet the statutory criteria. See Letter from Ronald Weich, Assistant Attorney General, to Joseph R. Biden, Jr., 1 (Apr. 30, 2012) (In 2011, of the 1,676 applications to the intelligence court, 2 were withdrawn by the Government, and the remaining 1,674 were approved, 30 with some modification), online at http://www.justice.gov/ nsd/foia/foia_library/2011fisa-ltr.pdf. (as visited Feb. 22, 2013, and available in Clerk of Court’s case file). As the intelligence court itself has stated, its review under § 1881a is “ ‘narrowly circumscribed.’ ” In re Proceedings Required by § 702(i) of the FISA Amendments Act of 2008, No. Mise. 08-01 (Aug. 27, 2008), p. 3. There is no reason to believe that the communications described would all fail to meet the conditions necessary for approval. Moreover, compared with prior law, § 1881a simplifies and thus expedites the approval process, making it more likely that the Government will use § 1881a to obtain the necessary approval.

The upshot is that (1) similarity of content, (2) strong motives, (3) prior behavior, and (4) capacity all point to a very strong likelihood that the Government will intercept at least some of the plaintiffs’ communications, including some that the 2008 amendment, § 1881a, but not the pre-2008 Act, authorizes the Government to intercept.

At the same time, nothing suggests the presence of some special factor here that might support a contrary conclusion. *431The Government does not deny that it has both the motive and the capacity to listen to communications of the kind described by the plaintiffs. Nor does it describe any system for avoiding the interception of an electronic communication that happens to include a party who is an American lawyer, journalist, or human rights worker. One can, of course, always imagine some special circumstance that negates a virtual likelihood, no matter how strong. But the same is true about most, if not all, ordinary inferences about future events. Perhaps, despite pouring rain, the streets will remain dry (due to the presence of a special chemical). But ordinarily a party that seeks to defeat a strong natural inference must bear the burden of showing that some such special circumstance exists. And no one has suggested any such special circumstance here.

Consequently, we need only assume that the Government is doing its job (to find out about, and combat, terrorism) in order to conclude that there is a high probability that the Government will intercept at least some electronic communication to which at least some of the plaintiffs are parties. The majority is wrong when it describes the harm threatened the plaintiffs as “speculative.”

IY

A

The majority more plausibly says that the plaintiffs have failed to show that the threatened harm is “certainly impending.” Ante, at 409 (internal quotation marks omitted). But, as the majority appears to concede, see ante, at 414, and n. 5, certainty is not, and never has been, the touchstone of standing. The future is inherently uncertain. Yet federal courts frequently entertain actions for injunctions and for declaratory relief aimed at preventing future activities that are reasonably likely or highly likely, but not absolutely certain, to take place. And that degree of certainty is all that is needed to support standing here.

*432The Court’s use of the term “certainly impending” is not to the contrary. Sometimes the Court has used the phrase “certainly impending” as if the phrase described a sufficient, rather than a necessary, condition for jurisdiction. See Pennsylvania v. West Virginia, 262 U. S. 553, 593 (1923) (“If the injury is certainly impending that is enough”). See also Babbitt, 442 U. S., at 298 (same). On other occasions, it has used the phrase as if it concerned when, not whether, an alleged injury would occur. Thus, in Lujan, 504 U. S., at 564, n. 2, the Court considered a threatened future injury that consisted of harm that the plaintiffs would suffer when they “soon” visited a government project area that (they claimed) would suffer environmental damage. The Court wrote that a “mere profession of an intent, some day, to return” to the project area did not show the harm was “imminent,” for “soon” might mean nothing more than “in this lifetime.” Id., at 564-565, n. 2 (internal quotation marks omitted). Similarly, in McConnell v. Federal Election Comm’n, 540 U. S. 93 (2003), the Court denied standing because the Senator’s future injury (stemming from a campaign finance law) would not affect him until his reelection. That fact, the Court said, made the injury “too remote temporally to satisfy Article III standing.” Id., at 225-226.

On still other occasions, recognizing that “‘imminence’ is concededly a somewhat elastic concept,” Lujan, supra, at 565, n. 2, the Court has referred to, or used (sometimes along with “certainly impending”), other phrases such as “reasonable probability” that suggest less than absolute, or literal, certainty. See Babbitt, supra, at 298 (plaintiff “must demonstrate a realistic danger of sustaining a direct injury” (emphasis added)); Friends of the Earth, Inc. v. Laidlaw Environmental Services (TOC), Inc., 528 U. S. 167, 190 (2000) (“[I]t is the plaintiff’s burden to establish standing by demonstrating that... the defendant’s allegedly wrongful behavior will likely occur or continue”). See also Monsanto Co. v. Geertson Seed Farms, 561 U. S. 139, 153 (2010) (“ ‘ “reason*433able probability” ’ ” and “substantial risk”); Davis, 554 U. S., at 734 (“realistic and impending threat of direct injury”); MedImmune, Inc. v. Genentech, Inc., 549 U. S. 118, 129 (2007) (“genuine threat of enforcement”); Department of Commerce v. United States House of Representatives, 525 U. S. 316, 333 (1999) (“substantially likely” (internal quotation marks omitted)); Clinton v. City of New York, 524 U. S. 417, 432 (1998) (“sufficient likelihood of economic injury”); Pennell v. San Jose, 485 U. S. 1, 8 (1988) (“realistic danger” (internal quotation marks omitted)); Blum v. Yaretsky, 457 U. S. 991, 1001 (1982) (“quite realistic” threat); Bryant v. Yellen, 447 U. S. 352, 367-368 (1980) (“likely”); Buckley v. Valeo, 424 U. S. 1, 74 (1976) (per curiam) (“reasonable probability”). Taken together the case law uses the word “certainly” as if it emphasizes, rather than literally defines, the immediately following term “impending.”

B

1

More important, the Court’s holdings in standing cases show that standing exists here. The Court has often found standing where the occurrence of the relevant injury was far less certain than here. Consider a few, fairly typical, cases. Consider Pennell, supra. A city ordinance forbade landlords to raise the rent charged to a tenant by more than 8 percent where doing so would work an unreasonably severe hardship on that tenant. Id., at 4-5. A group of landlords sought a judgment declaring the ordinance unconstitutional. The Court held that, to have standing, the landlords had to demonstrate a “ ‘realistic danger of sustaining a direct injury as a result of the statute’s operation.’ ” Id., at 8 (emphasis added). It found that the landlords had done so by showing a likelihood of enforcement and a “probability,” ibid., that the ordinance would make the landlords charge lower rents—even though the landlords had not shown (1) that they intended to raise the relevant rents to the point of *434causing unreasonably severe hardship; (2) that the tenants would challenge those increases; or (3) that the city’s hearing examiners and arbitrators would find against the landlords. Here, even more so than in Pennell, there is a “realistic danger” that the relevant harm will occur.

Or, consider Blum, supra. A group of nursing home residents receiving Medicaid benefits challenged the constitutionality (on procedural grounds) of a regulation that permitted their nursing home to transfer them to a less desirable home. Id., at 999-1000. Although a Medicaid committee had recommended transfers, Medicaid-initiated transfer had been enjoined and the nursing home itself had not threatened to transfer the plaintiffs. But the Court found “standing” because “the threat of transfers” was “not ‘imaginary or speculative’” but “quite realistic,” hence “sufficiently substantial.” Id., at 1000-1001 (quoting Younger v. Harris, 401 U. S. 37, 42 (1971)). The plaintiffs’ injury here is not imaginary or speculative, but “quite realistic.”

Or, consider Davis, supra. The plaintiff, a candidate for the United States House of Representatives, self-financed his campaigns. He challenged the constitutionality of an election law that relaxed the limits on an opponent’s contributions when a self-financed candidate’s spending itself exceeded certain other limits. His opponent, in fact, had decided not to take advantage of the increased contribution limits that the statute would have allowed. Id., at 734. But the Court nonetheless found standing because there was a “realistic and impending threat,” not a certainty, that the candidate’s opponent would do so at the time the plaintiff filed the complaint. Id., at 734-735. The threat facing the plaintiffs here is as “realistic and impending.”

Or, consider MedImmune, supra. The plaintiff, a patent licensee, sought a declaratory judgment that the patent was invalid. But, the plaintiff did not face an imminent threat of suit because it continued making royalty payments to the patent holder. In explaining why the plaintiff had standing, *435we (1) assumed that if the plaintiff stopped making royalty-payments it would have standing (despite the fact that the patent holder might not bring suit), (2) rejected the Federal Circuit’s “reasonable apprehension of imminent suit” requirement, and (3) instead suggested that a “genuine threat of enforcement” was likely sufficient. Id., at 128,129, 132, n. 11 (internal quotation marks omitted). A “genuine threat” is present here.

Moreover, courts have often found probabilistic injuries sufficient to support standing. In Duke Power Co. v. Carolina Environmental Study Group, Inc., 438 U. S. 59 (1978), for example, the plaintiffs, a group of individuals living near a proposed nuclear powerplant, challenged the constitutionality of the Price-Anderson Act, a statute that limited the plant’s liability in the case of a nuclear accident. The plaintiffs said that, without the Act, the defendants would not build a nuclear plant. And the building of the plant would harm them, in part, by emitting “non-natural radiation into [their] environment.” Id., at 74. The Court found standing in part due to “our generalized concern about exposure to radiation and the apprehension flowing from the uncertainty about the health and genetic consequences of even small emissions.” Ibid, (emphasis added). See also Monsanto Co., 561 U. S., at 153-154 (“A substantial risk of gene flow injures respondents in several ways” (emphasis added)).

See also lower court cases, such as Mountain States Legal Foundation v. Glickman, 92 F. 3d 1228, 1234-1235 (CADC 1996) (plaintiffs attack Government decision to limit timber harvesting; standing based upon increased risk of wildfires); Natural Resources Defense Council v. EPA, 464 F. 3d 1, 7 (CADC 2006) (plaintiffs attack Government decision deregulating methyl bromide; standing based upon increased lifetime risk of developing skin cancer); Constellation Energy Commodities Group, Inc. v. FERC, 457 F. 3d 14, 20 (CADC 2006) (standing based on increased risk of nonreeovery inherent in the reduction of collateral securing a debt of uncer*436tain amount); Sutton v. St. Jude Medical S. C., Inc., 419 F. 3d 568, 570-575 (CA6 2005) (standing based on increased risk of harm caused by implantation of defective medical device); Johnson v. Allsteel, Inc., 259 F. 3d 885, 888-891 (CA7 2001) (standing based on increased risk that Employee Retirement Income Security Act of 1974 beneficiary will not be covered due to increased amount of discretion given to ERISA administrator).

How could the law be otherwise? Suppose that a federal court faced a claim by homeowners that (allegedly) unlawful dam-building practices created a high risk that their homes would be flooded. Would the court deny them standing on the ground that the risk of flood was only 60, rather than 90, percent?

Would federal courts deny standing to a plaintiff in a diversity action who claims an anticipatory breach of contract where the future breach depends on probabilities? The defendant, say, has threatened to load wheat onto a ship bound for India despite a promise to send the wheat to the United States. No one can know for certain that this will happen. Perhaps the defendant will change his mind; perhaps the ship will turn and head for the United States. Yet, despite the uncertainty, the Constitution does not prohibit a federal court from hearing such a claim. See 23 R. Lord, Williston on Contracts §63:35 (4th ed. 2002) (plaintiff may bring an anticipatory breach suit even though the defendant’s promise is one to perform in the future, it has not yet been broken, and defendant may still retract the repudiation). E. g., Wisconsin Power & Light Co. v. Century Indemnity Co., 130 F. 3d 787, 792-793 (CA7 1997) (plaintiff could sue insurer that disclaimed liability for all costs that would be incurred in the future if environmental agencies required cleanup); Combs v. International Ins. Co., 354 F. 3d 568, 598-601 (CA6 2004) (similar).

Would federal courts deny standing to a plaintiff who seeks to enjoin as a nuisance the building of a nearby pond *437which, the plaintiff believes, will very likely, but not inevitably, overflow his land? See 42 Am. Jur. 2d, Injunctions §§2, 5 (2010) (noting that an injunction is ordinarily preventive in character and restrains actions that have not yet been taken, but threaten injury). E. g., Central Delta Water Agency v. United States, 306 F. 3d 938, 947-950 (CA9 2002) (standing to seek injunction where method of operating dam was highly likely to severely hamper plaintiffs’ ability to grow crops); Consolidated Companies, Inc. v. Union Pacific R. Co., 499 F. 3d 382, 386 (CA5 2007) (standing to seek injunction requiring cleanup of land adjacent to plaintiff’s tract because of threat that contaminants might migrate to plaintiff’s tract).

Neither do ordinary declaratory judgment actions always involve the degree of certainty upon which the Court insists here. See, e. g., Maryland Casualty Co. v. Pacific Coal & Oil Co., 312 U. S. 270, 273 (1941) (insurance company could seek declaration that it need not pay claim against insured automobile driver who was in an accident even though the driver had not yet been found liable for the accident); Aetna Life Ins. Co. v. Haworth, 300 U. S. 227, 239-244 (1937) (insurance company could seek declaration that it need not pay plaintiff for disability although plaintiff had not yet sought disability payments). See also, e. g., Associated Indemnity Corp. v. Fairchild Industries, Inc., 961 F. 2d 32, 35-36 (CA2 1992) (insured could seek declaration that insurance company must pay liability even before insured found liable).

2

In some standing cases, the Court has found that a reasonable probability of future injury comes accompanied with present injury that takes the form of reasonable efforts to mitigate the threatened effects of the future injury or to prevent it from occurring. Thus, in Monsanto Co., 561 U. S., at 153-156, the plaintiffs, a group of conventional alfalfa growers) challenged an agency decision to deregulate genetically *438engineered alfalfa. They claimed that deregulation would harm them because their neighbors would plant the genetically engineered seed, bees would obtain pollen from the neighbors’ plants, and the bees would then (harmfully) contaminate their own conventional alfalfa with the genetically modified gene. The lower courts had found a “reasonable probability” that this injury would occur. Ibid, (internal quotation marks omitted).

Without expressing views about that probability, we found standing because the plaintiffs would suffer present harm by trying to combat the threat. Ibid. The plaintiffs, for example, “would have to conduct testing to find out whether and to what extent their crops have been contaminated.” Id., at 154. And they would have to take “measures to minimize the likelihood of potential contamination and to ensure an adequate supply of non-genetically-engineered alfalfa.” Ibid. We held that these “harms, which [the plaintiffs] will suffer even if their crops are not actually infected with” the genetically modified gene, “are sufficiently concrete to satisfy the injury-in-fact prong of the constitutional standing analysis.” Id., at 155.

Virtually identical circumstances are present here. Plaintiff McKay, for example, points out that, when he communicates abroad about, or in the interests of, a client (e. g., a client accused of terrorism), he must “make an assessment” whether his “client’s interest would be compromised” should the Government “acquire the communications.” App. to Pet. for Cert. 375a. If so, he must either forgo the communication or travel abroad. Id., at 371a-372a (“I have had to take measures to protect the confidentiality of information that I believe is particularly sensitive,” including “travel that is both time-consuming and expensive”).

Since travel is expensive, since forgoing communication can compromise the client’s interests, since McKay’s assessment itself takes time and effort, this case does not differ significantly from Monsanto. And that is so whether we *439consider the plaintiffs’ present necessary expenditure of time and effort as a separate concrete, particularized, imminent harm, or consider it as additional evidence that the future harm (an interception) is likely, to occur. See also Friends of the Earth, Inc., 528 U. S., at 183-184 (holding that plaintiffs who curtailed their recreational activities on a river due to reasonable concerns about the effect of pollutant discharges into that river had standing); Meese v. Keene, 481 U. S. 465, 475 (1987) (stating that “the need to take ... affirmative steps to avoid the risk of harm . . . constitutes a cognizable injury”).

3

The majority cannot find support in cases that use the words “certainly impending” to deny standing. While I do not claim to have read every standing case, I have examined quite a few, and not yet found any such case. The majority refers to Whitmore v. Arkansas, 495 U. S. 149 (1990). But in that case the Court denied standing to a prisoner who challenged the validity of a death sentence given to a different prisoner who refused to challenge his own sentence. The plaintiff feared that in the absence of an appeal, his fellow prisoner’s death sentence would be missing from the State’s death penalty database and thereby skew the database against him, making it less likely his challenges to his own death penalty would succeed. The Court found no standing. Id., at 161. But the fellow prisoner’s lack of appeal would have harmed the plaintiff only if (1) the plaintiff separately obtained federal habeas relief and was then recon-victed and resentenced to death, (2) he sought review of his new sentence, and (3) during that review, his death sentence was affirmed only because it was compared to an artificially skewed database. Id., at 156-157. These events seemed not very likely to occur.

In Daimler Chrysler Corp. v. Cuno, 547 U. S. 332 (2006), taxpayers challenged the constitutionality of a tax break offered by state and local governments to a car manufacturer. *440We found no standing. But the plaintiffs would have suffered resulting injury only if the tax break had depleted state and local treasuries and the legislature had responded by raising their taxes. Id., at 344.

In Lujan, the case that may come closest to supporting the majority, the Court also found no standing. But, as I pointed out, supra, at 432, Lujan is a case where the Court considered when, not whether, the threatened harm would occur. 504 U. S., at 564, n. 2. The relevant injury there consisted of a visit by an environmental group’s members to a project site where they would find (unlawful) environmental depredation. Id., at 564. The Court pointed out that members had alleged that they would visit the project sites “soon.” But it wrote that “soon” might refer to almost any time in the future. Ibid., n. 2. By way of contrast, the ongoing threat of terrorism means that here the relevant interceptions will likely take place imminently, if not now.

The Court has, of course, denied standing in other cases. But they involve injuries less likely, not more likely, to occur than here. In a recent case, Summers v. Earth Island Institute, 555 U. S. 488 (2009), for example, the plaintiffs challenged a regulation exempting certain timber sales from public comment and administrative appeal. The plaintiffs claimed that the regulations injured them by interfering with their esthetic enjoyment and recreational use of the forests. The Court found this harm too unlikely to occur to support standing. Id., at 496. The Court noted that one plaintiff had not pointed to a specific affected forest that he would visit. The Court concluded that “[tjhere may be a chance, but . . . hardly a likelihood,” that the plaintiff’s “wanderings will bring him to a parcel about to be affected by a project unlawfully subject to the regulations.” Id., at 495 (emphasis added).

4

In sum, as the Court concedes, see ante, at 414, and n. 5, the word “certainly” in the phrase “certainly impending” *441does not refer to absolute certainty. As our case law demonstrates, what the Constitution requires is something more akin to “reasonable probability” or “high probability.” The use of some such standard is all that is necessary here to ensure the actual concrete injury that the Constitution demands. The considerations set forth in Parts II and III, supra, make clear that the standard is readily met in this case.

⅜ * *

While I express no view on the merits of the plaintiffs’ constitutional claims, I do believe that at least some of the plaintiffs have standing to make those claims. I dissent, with respect, from the majority’s contrary conclusion.

Chief Justice ROBERTS delivered the opinion of the Court.

*2561The Secretary of Commerce decided to reinstate a question about citizenship on the 2020 census questionnaire. A group of plaintiffs challenged that decision on constitutional and statutory grounds. We now decide whether the Secretary violated the Enumeration Clause of the Constitution, the Census Act, or otherwise abused his discretion.

I

A

In order to apportion Members of the House of Representatives among the States, the Constitution requires an "Enumeration" of the population every 10 years, to be made "in such Manner" as Congress "shall by Law direct." Art. I, § 2, cl. 3; Amdt. 14, § 2. In the Census Act, Congress delegated to the Secretary of Commerce the task of conducting the decennial census "in such form and content as he may determine." 13 U. S. C. § 141(a). The Secretary is aided in that task by the Census Bureau, a statistical agency housed within the Department of Commerce. See §§ 2, 21.

The population count derived from the census is used not only to apportion representatives but also to allocate federal funds to the States and to draw electoral districts. Wisconsin v. City of New York , 517 U.S. 1, 5-6, 116 S.Ct. 1091, 134 L.Ed.2d 167 (1996). The census additionally serves as a means of collecting demographic information, which "is used for such varied purposes as computing federal grant-in-aid benefits, drafting of legislation, urban and regional planning, business planning, and academic and social studies." Baldrige v. Shapiro , 455 U.S. 345, 353-354, n. 9, 102 S.Ct. 1103, 71 L.Ed.2d 199 (1982). Over the years, the census has asked questions about (for example) race, sex, age, health, education, occupation, housing, and military service. It has also asked about radio ownership, age at first marriage, and native tongue. The Census Act obliges everyone to answer census questions truthfully and requires the Secretary to keep individual answers confidential, including from other Government agencies. §§ 221, 8(b), 9(a).

There have been 23 decennial censuses from the first census in 1790 to the most recent in 2010. Every census between 1820 and 2000 (with the exception of 1840) asked at least some of the population about their citizenship or place of birth. Between 1820 and 1950, the question was asked of all households. Between 1960 and 2000, it was asked of about one-fourth to one-sixth of the population. That change was part of a larger effort to simplify the census by asking most people a few basic demographic questions (such as sex, age, race, and marital status) on a short-form questionnaire, while asking a sample of the population more detailed demographic questions on a long-form questionnaire. In explaining the decision to move the citizenship question to the long-form questionnaire, the Census Bureau opined that "general census information on citizenship had become of less importance compared with other possible questions to be included in the census, particularly in view of the *2562recent statutory requirement for annual alien registration which could provide the Immigration and Naturalization Service, the principal user of such data, with the information it needed." Dept. of Commerce, Bureau of Census, 1960 Censuses of Population and Housing 194 (1966).1

In 2010, the year of the latest census, the format changed again. All households received the same questionnaire, which asked about sex, age, race, Hispanic origin, and living arrangements. The more detailed demographic questions previously asked on the long-form questionnaire, including the question about citizenship, were instead asked in the American Community Survey (or ACS), which is sent each year to a rotating sample of about 2.6% of households.

The Census Bureau and former Bureau officials have resisted occasional proposals to resume asking a citizenship question of everyone, on the ground that doing so would discourage noncitizens from responding to the census and lead to a less accurate count of the total population. See, e.g. , Federation of Am. Immigration Reform v. Klutznick , 486 F.Supp. 564, 568 (DDC 1980) ("[A]ccording to the Bureau[,] any effort to ascertain citizenship will inevitably jeopardize the overall accuracy of the population count"); Brief for Former Directors of the U. S. Census Bureau as Amici Curiae in Evenwel v. Abbott , O. T. 2014, No. 14-940, p. 25 (inquiring about citizenship would "invariably lead to a lower response rate").

B

In March 2018, Secretary of Commerce Wilbur Ross announced in a memo that he had decided to reinstate a question about citizenship on the 2020 decennial census questionnaire. The Secretary stated that he was acting at the request of the Department of Justice (DOJ), which sought improved data about citizen voting-age population for purposes of enforcing the Voting Rights Act (or VRA)-specifically the Act's ban on diluting the influence of minority voters by depriving them of single-member districts in which they can elect their preferred candidates. App. to Pet. for Cert. 548a. DOJ explained that federal courts determine whether a minority group could constitute a majority in a particular district by looking to the citizen voting-age population of the group. According to DOJ, the existing citizenship data from the American Community Survey was not ideal: It was not reported at the level of the census block, the basic component of legislative districting plans; it had substantial margins of error; and it did not align in time with the census-based population counts used to draw legislative districts. DOJ therefore formally requested reinstatement of the citizenship question on the census questionnaire. Id. , at 565a-569a.

The Secretary's memo explained that the Census Bureau initially analyzed, and the Secretary considered, three possible courses of action. The first was to continue to collect citizenship information in the American Community Survey and attempt to develop a data model that would more accurately estimate citizenship at the census block level. The Secretary rejected that option because the Bureau "did not assert and could not confirm" that such ACS-based data modeling was possible "with a sufficient degree of accuracy." Id. , at 551a.

*2563The second option was to reinstate a citizenship question on the decennial census. The Bureau predicted that doing so would discourage some noncitizens from responding to the census. That would necessitate increased "non-response follow up" operations-procedures the Bureau uses to attempt to count people who have not responded to the census-and potentially lead to a less accurate count of the total population.

Option three was to use administrative records from other agencies, such as the Social Security Administration and Citizenship and Immigration Services, to provide DOJ with citizenship data. The Census Bureau recommended this option, and the Secretary found it a "potentially appealing solution" because the Bureau has long used administrative records to supplement and improve census data. Id. , at 554a. But the Secretary concluded that administrative records alone were inadequate because they were missing for more than 10% of the population.

The Secretary ultimately asked the Census Bureau to develop a fourth option that would combine options two and three: reinstate a citizenship question on the census questionnaire, and also use the time remaining until the 2020 census to "further enhance" the Bureau's "administrative record data sets, protocols, and statistical models." Id. , at 555a. The memo explained that, in the Secretary's judgment, the fourth option would provide DOJ with the "most complete and accurate" citizen voting-age population data in response to its request. Id. , at 556a.

The Secretary "carefully considered" the possibility that reinstating a citizenship question would depress the response rate. Ibid . But after evaluating the Bureau's "limited empirical evidence" on the question-evidence drawn from estimated non-response rates to previous American Community Surveys and census questionnaires-the Secretary concluded that it was not possible to "determine definitively" whether inquiring about citizenship in the census would materially affect response rates. Id. , at 557a, 562a. He also noted the long history of the citizenship question on the census, as well as the facts that the United Nations recommends collecting census-based citizenship information, and other major democracies such as Australia, Canada, France, Indonesia, Ireland, Germany, Mexico, Spain, and the United Kingdom inquire about citizenship in their censuses. Altogether, the Secretary determined that "the need for accurate citizenship data and the limited burden that the reinstatement of the citizenship question would impose outweigh fears about a potentially lower response rate." Id. , at 557a.

C

Shortly after the Secretary announced his decision, two groups of plaintiffs filed suit in Federal District Court in New York, challenging the decision on several grounds. The first group of plaintiffs included 18 States, the District of Columbia, various counties and cities, and the United States Conference of Mayors. They alleged that the Secretary's decision violated the Enumeration Clause of the Constitution and the requirements of the Administrative Procedure Act. The second group of plaintiffs consisted of several non-governmental organizations that work with immigrant and minority communities. They added an equal protection claim. The District Court consolidated the two cases. Both groups of plaintiffs are respondents here.

The Government moved to dismiss the lawsuits, arguing that the Secretary's decision was unreviewable and that respondents had failed to state cognizable claims *2564under the Enumeration Clause and the Equal Protection Clause. The District Court dismissed the Enumeration Clause claim but allowed the other claims to proceed. 315 F.Supp.3d 766 (SDNY 2018).

In June 2018, the Government submitted to the District Court the Commerce Department's "administrative record": the materials that Secretary Ross considered in making his decision. That record included DOJ's December 2017 letter requesting reinstatement of the citizenship question, as well as several memos from the Census Bureau analyzing the predicted effects of reinstating the question. Shortly thereafter, at DOJ's urging, the Government supplemented the record with a new memo from the Secretary, "intended to provide further background and context regarding" his March 2018 memo. App. to Pet. for Cert. 546a. The supplemental memo stated that the Secretary had begun considering whether to add the citizenship question in early 2017, and had inquired whether DOJ "would support, and if so would request, inclusion of a citizenship question as consistent with and useful for enforcement of the Voting Rights Act." Ibid. According to the Secretary, DOJ "formally" requested reinstatement of the citizenship question after that inquiry. Ibid.

Respondents argued that the supplemental memo indicated that the Government had submitted an incomplete record of the materials considered by the Secretary. They asked the District Court to compel the Government to complete the administrative record. The court granted that request, and the parties jointly stipulated to the inclusion of more than 12,000 pages of additional materials in the administrative record. Among those materials were emails and other records confirming that the Secretary and his staff began exploring the possibility of reinstating a citizenship question shortly after he was confirmed in early 2017, attempted to elicit requests for citizenship data from other agencies, and eventually persuaded DOJ to request reinstatement of the question for VRA enforcement purposes.

In addition, respondents asked the court to authorize discovery outside the administrative record. They claimed that such an unusual step was warranted because they had made a strong preliminary showing that the Secretary had acted in bad faith. See Citizens to Preserve Overton Park, Inc. v. Volpe , 401 U.S. 402, 420, 91 S.Ct. 814, 28 L.Ed.2d 136 (1971). The court also granted that request, authorizing expert discovery and depositions of certain DOJ and Commerce Department officials.

In August and September 2018, the District Court issued orders compelling depositions of Secretary Ross and of the Acting Assistant Attorney General for DOJ's Civil Rights Division. We granted the Government's request to stay the Secretary's deposition pending further review, but we declined to stay the Acting AAG's deposition or the other extra-record discovery that the District Court had authorized.

The District Court held a bench trial and issued findings of fact and conclusions of law on respondents' statutory and equal protection claims. After determining that respondents had standing to sue, the District Court ruled that the Secretary's action was arbitrary and capricious, based on a pretextual rationale, and violated certain provisions of the Census Act. On the equal protection claim, however, the District Court concluded that respondents had not met their burden of showing that the Secretary was motivated by discriminatory animus. The court granted judgment to respondents on their statutory claims, vacated the Secretary's decision, and enjoined him from reinstating the citizenship question until he cured the legal errors the *2565court had identified. 351 F.Supp.3d 502 (SDNY 2019).

The Government appealed to the Second Circuit, but also filed a petition for writ of certiorari before judgment, asking this Court to review the District Court's decision directly because the case involved an issue of imperative public importance, and the census questionnaire needed to be finalized for printing by the end of June 2019. We granted the petition. 586 U. S. ----, 139 S.Ct. 16, 202 L.Ed.2d 306 (2019). At the Government's request, we later ordered the parties to address whether the Enumeration Clause provided an alternative basis to affirm. 586 U. S. ----, 139 S.Ct. 16, 202 L.Ed.2d 306 (2019).

II

We begin with jurisdiction. Article III of the Constitution limits federal courts to deciding "Cases" and "Controversies." For a legal dispute to qualify as a genuine case or controversy, at least one plaintiff must have standing to sue. The doctrine of standing "limits the category of litigants empowered to maintain a lawsuit in federal court to seek redress for a legal wrong" and "confines the federal courts to a properly judicial role." Spokeo , Inc. v. Robins , 578 U. S. ----, ----, 136 S.Ct. 1540, 1547, 194 L.Ed.2d 635 (2016). To have standing, a plaintiff must "present an injury that is concrete, particularized, and actual or imminent; fairly traceable to the defendant's challenged behavior; and likely to be redressed by a favorable ruling." Davis v. Federal Election Comm'n , 554 U.S. 724, 733, 128 S.Ct. 2759, 171 L.Ed.2d 737 (2008).

Respondents assert a number of injuries-diminishment of political representation, loss of federal funds, degradation of census data, and diversion of resources-all of which turn on their expectation that reinstating a citizenship question will depress the census response rate and lead to an inaccurate population count. Several States with a disproportionate share of noncitizens, for example, anticipate losing a seat in Congress or qualifying for less federal funding if their populations are undercounted. These are primarily future injuries, which "may suffice if the threatened injury is certainly impending, or there is a substantial risk that the harm will occur." Susan B. Anthony List v. Driehaus , 573 U.S. 149, 158, 134 S.Ct. 2334, 189 L.Ed.2d 246 (2014) (internal quotation marks omitted).

The District Court concluded that the evidence at trial established a sufficient likelihood that the reinstatement of a citizenship question would result in noncitizen households responding to the census at lower rates than other groups, which in turn would cause them to be undercounted and lead to many of respondents' asserted injuries. For purposes of standing, these findings of fact were not so suspect as to be clearly erroneous.

We therefore agree that at least some respondents have Article III standing. Several state respondents here have shown that if noncitizen households are undercounted by as little as 2%-lower than the District Court's 5.8% prediction-they will lose out on federal funds that are distributed on the basis of state population. That is a sufficiently concrete and imminent injury to satisfy Article III, and there is no dispute that a ruling in favor of respondents would redress that harm.

The Government contends, however, that any harm to respondents is not fairly traceable to the Secretary's decision, because such harm depends on the independent action of third parties choosing to violate their legal duty to respond to the census. The chain of causation is made *2566even more tenuous, the Government argues, by the fact that such intervening, unlawful third-party action would be motivated by unfounded fears that the Federal Government will itself break the law by using noncitizens' answers against them for law enforcement purposes. The Government invokes our steady refusal to "endorse standing theories that rest on speculation about the decisions of independent actors," Clapper v. Amnesty Int'l USA , 568 U.S. 398, 414, 133 S.Ct. 1138, 185 L.Ed.2d 264 (2013), particularly speculation about future unlawful conduct, Los Angeles v. Lyons , 461 U.S. 95, 105, 103 S.Ct. 1660, 75 L.Ed.2d 675 (1983).

But we are satisfied that, in these circumstances, respondents have met their burden of showing that third parties will likely react in predictable ways to the citizenship question, even if they do so unlawfully and despite the requirement that the Government keep individual answers confidential. The evidence at trial established that noncitizen households have historically responded to the census at lower rates than other groups, and the District Court did not clearly err in crediting the Census Bureau's theory that the discrepancy is likely attributable at least in part to noncitizens' reluctance to answer a citizenship question. Respondents' theory of standing thus does not rest on mere speculation about the decisions of third parties; it relies instead on the predictable effect of Government action on the decisions of third parties. Cf. Bennett v. Spear , 520 U.S. 154, 169-170, 117 S.Ct. 1154, 137 L.Ed.2d 281 (1997) ; Davis , 554 U.S. at 734-735, 128 S.Ct. 2759. Because Article III "requires no more than de facto causality," Block v. Meese , 793 F.2d 1303, 1309 (CADC 1986) (Scalia, J.), traceability is satisfied here. We may therefore consider the merits of respondents' claims, at least as far as the Constitution is concerned.

III

The Enumeration Clause of the Constitution does not provide a basis to set aside the Secretary's decision. The text of that clause "vests Congress with virtually unlimited discretion in conducting the decennial 'actual Enumeration,' " and Congress "has delegated its broad authority over the census to the Secretary." Wisconsin , 517 U.S. at 19, 116 S.Ct. 1091. Given that expansive grant of authority, we have rejected challenges to the conduct of the census where the Secretary's decisions bore a "reasonable relationship to the accomplishment of an actual enumeration." Id. , at 20, 116 S.Ct. 1091.

Respondents ask us to evaluate the Secretary's decision to reinstate a citizenship question under that "reasonable relationship" standard, but we agree with the District Court that a different analysis is needed here. Our cases applying that standard concerned decisions about the population count itself-such as a postcensus decision not to use a particular method to adjust an undercount, id. , at 4, 116 S.Ct. 1091, and a decision to allocate overseas military personnel to their home States, Franklin v. Massachusetts , 505 U.S. 788, 790-791, 112 S.Ct. 2767, 120 L.Ed.2d 636 (1992). We have never applied the standard to decisions about what kinds of demographic information to collect in the course of taking the census. Indeed, as the District Court recognized, applying the "reasonable relationship" standard to every census-related decision "would lead to the conclusion that it is unconstitutional to ask any demographic question on the census" because "asking such questions bears no relationship whatsoever to the goal of an accurate headcount." 315 F.Supp.3d at 804-805. Yet demographic questions have been asked in every census since 1790, and questions about citizenship in particular *2567have been asked for nearly as long. Like the District Court, we decline respondents' invitation to measure the constitutionality of the citizenship question by a standard that would seem to render every census since 1790 unconstitutional.

We look instead to Congress's broad authority over the census, as informed by long and consistent historical practice. All three branches of Government have understood the Constitution to allow Congress, and by extension the Secretary, to use the census for more than simply counting the population. Since 1790, Congress has sought, or permitted the Secretary to seek, information about matters as varied as age, sex, marital status, health, trade, profession, literacy, and value of real estate owned. See id. , at 801. Since 1820, it has sought, or permitted the Secretary to seek, information about citizenship in particular. Federal courts have approved the practice of collecting demographic data in the census. See, e.g. , United States v. Moriarity , 106 F. 886, 891 (CC SDNY 1901) (duty to take a census of population "does not prohibit the gathering of other statistics, if 'necessary and proper,' for the intelligent exercise of other powers enumerated in the constitution"). While we have never faced the question directly, we have assumed that Congress has the power to use the census for information-gathering purposes, see Legal Tender Cases , 12 Wall. 457, 536, 20 L.Ed. 287 (1871), and we have recognized the role of the census as a "linchpin of the federal statistical system by collecting data on the characteristics of individuals, households, and housing units throughout the country," Department of Commerce v. United States House of Representatives , 525 U.S. 316, 341, 119 S.Ct. 765, 142 L.Ed.2d 797 (1999) (internal quotation marks omitted).

That history matters. Here, as in other areas, our interpretation of the Constitution is guided by a Government practice that "has been open, widespread, and unchallenged since the early days of the Republic." NLRB v. Noel Canning , 573 U.S. 513, 572, 134 S.Ct. 2550, 189 L.Ed.2d 538 (2014) (Scalia, J., concurring in judgment); see also Wisconsin , 517 U.S. at 21, 116 S.Ct. 1091 (noting "importance of historical practice" in census context). In light of the early understanding of and long practice under the Enumeration Clause, we conclude that it permits Congress, and by extension the Secretary, to inquire about citizenship on the census questionnaire. We need not, and do not, decide the constitutionality of any other question that Congress or the Secretary might decide to include in the census.

IV

The District Court set aside the Secretary's decision to reinstate a citizenship question on the grounds that the Secretary acted arbitrarily and violated certain provisions of the Census Act. The Government contests those rulings, but also argues that the Secretary's decision was not judicially reviewable under the Administrative Procedure Act in the first place. We begin with that contention.

A

The Administrative Procedure Act embodies a "basic presumption of judicial review," Abbott Laboratories v. Gardner , 387 U.S. 136, 140, 87 S.Ct. 1507, 18 L.Ed.2d 681 (1967), and instructs reviewing courts to set aside agency action that is "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law," 5 U. S. C. § 706(2)(A). Review is not available, however, "to the extent that" a relevant statute precludes it, § 701(a)(1), or the agency action is "committed to agency discretion by law," § 701(a)(2). The Government argues that the Census Act *2568commits to the Secretary's unreviewable discretion decisions about what questions to include on the decennial census questionnaire.

We disagree. To be sure, the Act confers broad authority on the Secretary. Section 141(a) instructs him to take "a decennial census of population" in "such form and content as he may determine, including the use of sampling procedures and special surveys." 13 U. S. C. § 141. The Act defines "census of population" to mean "a census of population, housing, and matters relating to population and housing," § 141(g), and it authorizes the Secretary, in "connection with any such census," to "obtain such other census information as necessary," § 141(a). It also states that the "Secretary shall prepare questionnaires, and shall determine the inquiries, and the number, form, and subdivisions thereof, for the statistics, surveys, and censuses provided for in this title." § 5. And it authorizes him to acquire materials, such as administrative records, from other federal, state, and local agencies in aid of conducting the census. § 6. Those provisions leave much to the Secretary's discretion. See Wisconsin , 517 U.S. at 19, 116 S.Ct. 1091 ("Through the Census Act, Congress has delegated its broad authority over the census to the Secretary.").

But they do not leave his discretion unbounded. In order to give effect to the command that courts set aside agency action that is an abuse of discretion, and to honor the presumption of judicial review, we have read the § 701(a)(2) exception for action committed to agency discretion "quite narrowly, restricting it to 'those rare circumstances where the relevant statute is drawn so that a court would have no meaningful standard against which to judge the agency's exercise of discretion.' " Weyerhaeuser Co. v. United States Fish and Wildlife Serv. , 586 U. S. ----, ----, 139 S.Ct. 361, 370, 202 L.Ed.2d 269 (2018) (quoting Lincoln v. Vigil , 508 U.S. 182, 191, 113 S.Ct. 2024, 124 L.Ed.2d 101 (1993) ). And we have generally limited the exception to "certain categories of administrative decisions that courts traditionally have regarded as 'committed to agency discretion,' " id. , at 191, 113 S.Ct. 2024, such as a decision not to institute enforcement proceedings, Heckler v. Chaney , 470 U.S. 821, 831-832, 105 S.Ct. 1649, 84 L.Ed.2d 714 (1985), or a decision by an intelligence agency to terminate an employee in the interest of national security, Webster v. Doe , 486 U.S. 592, 600-601, 108 S.Ct. 2047, 100 L.Ed.2d 632 (1988).

The taking of the census is not one of those areas traditionally committed to agency discretion. We and other courts have entertained both constitutional and statutory challenges to census-related decisionmaking. See, e.g. , Department of Commerce , 525 U.S. 316, 119 S.Ct. 765, 142 L.Ed.2d 797 ; Wisconsin , 517 U.S. 1, 116 S.Ct. 1091, 134 L.Ed.2d 167 ; Carey v. Klutznick , 637 F.2d 834 (CA2 1980).

Nor is the statute here drawn so that it furnishes no meaningful standard by which to judge the Secretary's action. In contrast to the National Security Act in Webster , which gave the Director of Central Intelligence discretion to terminate employees whenever he "deem[ed]" it "advisable," 486 U.S. at 594, 108 S.Ct. 2047, the Census Act constrains the Secretary's authority to determine the form and content of the census in a number of ways. Section 195, for example, governs the extent to which he can use statistical sampling. Section 6(c), which will be considered in more detail below, circumscribes his power in certain circumstances to collect information through direct inquiries when administrative records are available. More generally, by mandating a population count that will be used to apportion representatives, *2569see § 141(b), 2 U. S. C. § 2a, the Act imposes "a duty to conduct a census that is accurate and that fairly accounts for the crucial representational rights that depend on the census and the apportionment." Franklin , 505 U.S. at 819-820, 112 S.Ct. 2767 (Stevens, J., concurring in part and concurring in judgment).

The Secretary's decision to reinstate a citizenship question is amenable to review for compliance with those and other provisions of the Census Act, according to the general requirements of reasoned agency decisionmaking. Because this is not a case in which there is "no law to apply," Overton Park , 401 U.S. at 410, 91 S.Ct. 814, the Secretary's decision is subject to judicial review.

B

At the heart of this suit is respondents' claim that the Secretary abused his discretion in deciding to reinstate a citizenship question. We review the Secretary's exercise of discretion under the deferential "arbitrary and capricious" standard. See 5 U. S. C. § 706(2)(A). Our scope of review is "narrow": we determine only whether the Secretary examined "the relevant data" and articulated "a satisfactory explanation" for his decision, "including a rational connection between the facts found and the choice made." Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co. , 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983) (internal quotation marks omitted). We may not substitute our judgment for that of the Secretary, ibid. , but instead must confine ourselves to ensuring that he remained "within the bounds of reasoned decisionmaking," Baltimore Gas & Elec. Co. v. Natural Resources Defense Council, Inc. , 462 U.S. 87, 105, 103 S.Ct. 2246, 76 L.Ed.2d 437 (1983).

The District Court set aside the Secretary's decision for two independent reasons: His course of action was not supported by the evidence before him, and his stated rationale was pretextual. We focus on the first point here and take up the question of pretext later.

The Secretary examined the Bureau's analysis of various ways to collect improved citizenship data and explained why he thought the best course was to both reinstate a citizenship question and use citizenship data from administrative records to fill in the gaps. He considered but rejected the Bureau's recommendation to use administrative records alone. As he explained, records are lacking for about 10% of the population, so the Bureau would still need to estimate citizenship for millions of voting-age people. Asking a citizenship question of everyone, the Secretary reasoned, would eliminate the need to estimate citizenship for many of those people. And supplementing census responses with administrative record data would help complete the picture and allow the Bureau to better estimate citizenship for the smaller set of cases where it was still necessary to do so.

The evidence before the Secretary supported that decision. As the Bureau acknowledged, each approach-using administrative records alone, or asking about citizenship and using records to fill in the gaps-entailed tradeoffs between accuracy and completeness. Without a citizenship question, the Bureau would need to estimate the citizenship of about 35 million people; with a citizenship question, it would need to estimate the citizenship of only 13.8 million. Under either approach, there would be some errors in both the administrative records and the Bureau's estimates. With a citizenship question, there would also be some erroneous self-responses (about 500,000) and some conflicts *2570between responses and administrative record data (about 9.5 million).

The Bureau explained that the "relative quality" of the citizenship data generated by each approach would depend on the "relative importance of the errors" in each, but it was not able to "quantify the relative magnitude of the errors across the alternatives." App. 148. The Bureau nonetheless recommended using administrative records alone because it had "high confidence" that it could develop an accurate model for estimating the citizenship of the 35 million people for whom administrative records were not available, and it thought the resulting citizenship data would be of superior quality. Id. , at 146, 158-159. But when the time came for the Secretary to make a decision, the model did not yet exist, and even if it had, there was no way to gauge its relative accuracy. As the Bureau put it, "we will most likely never possess a fully adequate truth deck to benchmark" the model-which appears to be bureaucratese for "maybe, maybe not." Id. , at 146. The Secretary opted instead for the approach that would yield a more complete set of data at an acceptable rate of accuracy, and would require estimating the citizenship of fewer people.

The District Court overruled that choice, agreeing with the Bureau's assessment that its recommended approach would yield higher quality citizenship data on the whole. But the choice between reasonable policy alternatives in the face of uncertainty was the Secretary's to make. He considered the relevant factors, weighed risks and benefits, and articulated a satisfactory explanation for his decision. In overriding that reasonable exercise of discretion, the court improperly substituted its judgment for that of the agency.

The Secretary then weighed the benefit of collecting more complete and accurate citizenship data against the risk that inquiring about citizenship would depress census response rates, particularly among noncitizen households. In the Secretary's view, that risk was difficult to assess. The Bureau predicted a 5.1% decline in response rates among noncitizen households if the citizenship question were reinstated.2 It relied for that prediction primarily on studies showing that, while noncitizens had responded at lower rates than citizens to the 2000 short-form and 2010 censuses, which did not ask about citizenship, they responded at even lower rates than citizens to the 2000 long-form census and the 2010 American Community Survey, which did ask about citizenship. The Bureau thought it was reasonable to infer that the citizenship question accounted for the differential decline in noncitizen responses. But, the Secretary explained, the Bureau was unable to rule out other causes. For one thing, the evidence before the Secretary suggested that noncitizen households tend to be more distrustful of, and less likely to respond to, any government effort to collect information. For another, both the 2000 long-form census and 2010 ACS asked over 45 questions on a range of topics, including employment, income, and housing characteristics. Noncitizen households might disproportionately fail to respond to a lengthy and intrusive Government questionnaire for a number of reasons besides reluctance to answer a citizenship question-reasons relating to education level, socioeconomic status, and less exposure to Government outreach efforts. See App. to Pet. for Cert. 553a-554a, 557a-558a.

*2571The Secretary justifiably found the Bureau's analysis inconclusive. Weighing that uncertainty against the value of obtaining more complete and accurate citizenship data, he determined that reinstating a citizenship question was worth the risk of a potentially lower response rate. That decision was reasonable and reasonably explained, particularly in light of the long history of the citizenship question on the census.

Justice BREYER would conclude otherwise, but only by subordinating the Secretary's policymaking discretion to the Bureau's technocratic expertise. Justice BREYER's analysis treats the Bureau's (pessimistic) prediction about response rates and (optimistic) assumptions about its data modeling abilities as touchstones of substantive reasonableness rather than simply evidence for the Secretary to consider. He suggests that the Secretary should have deferred to the Bureau or at least offered some special justification for drawing his own inferences and adopting his own assumptions. But the Census Act authorizes the Secretary, not the Bureau, to make policy choices within the range of reasonable options. And the evidence before the Secretary hardly led ineluctably to just one reasonable course of action. It called for value-laden decisionmaking and the weighing of incommensurables under conditions of uncertainty. The Secretary was required to consider the evidence and give reasons for his chosen course of action. He did so. It is not for us to ask whether his decision was "the best one possible" or even whether it was "better than the alternatives." FERC v. Electric Power Supply Assn. , 577 U. S. ----, ----, 136 S.Ct. 760, 782, 193 L.Ed.2d 661 (2016). By second-guessing the Secretary's weighing of risks and benefits and penalizing him for departing from the Bureau's inferences and assumptions, Justice BREYER-like the District Court-substitutes his judgment for that of the agency.

C

The District Court also ruled that the Secretary violated two particular provisions of the Census Act, § 6(c) and § 141(f).

Section 6 has three subsections. Subsections (a) and (b) authorize the Secretary to acquire administrative records from other federal agencies and from state and local governments.3 Subsection (c) states:

"To the maximum extent possible and consistent with the kind, timeliness, quality and scope of the statistics required, the Secretary shall acquire and use information available from any source referred to in subsection (a) or (b) of this section instead of conducting direct inquiries." 13 U. S. C. § 6(c).

The District Court held, and respondents argue, that the Secretary failed to comply with § 6(c) because he opted to collect citizenship data using direct inquiries when it was possible to provide DOJ with data from administrative records alone.

*2572At the outset, § 6(c) may not even apply here. It governs the Secretary's choices with respect to "statistics required." The parties have assumed that phrase refers to census-related data that the Secretary wishes to acquire, but it may instead refer to particular kinds of statistics that other provisions of the Census Act actually do require the Secretary to collect and publish. See, e.g. , § 41 ("The Secretary shall collect and publish statistics concerning [cotton and cotton production]."); § 61 ("The Secretary shall collect, collate, and publish monthly statistics concerning [vegetable and animal oils and the like]."); § 91 ("The Secretary shall collect and publish quarterly financial statistics of business operations, organization, practices, management, and relation to other businesses."). If so, § 6(c) would seem to have nothing to say about the Secretary's collection of census-related citizenship data, which is not a "statistic" he is "required" to collect.

Regardless, assuming the provision applies, the Secretary complied with it, for essentially the same reasons that his decision was not arbitrary and capricious. As he explained, administrative records would not, in his judgment, provide the more complete and accurate data that DOJ sought. He thus could not, "consistent with" the kind and quality of the "statistics required," use administrative records instead of asking about citizenship directly. Respondents' arguments to the contrary rehash their disagreement with the Secretary's policy judgment about which approach would yield the most complete and accurate citizenship data. For the reasons already discussed, we may not substitute our judgment for that of the Secretary here.

We turn now to § 141(f), which requires the Secretary to report to Congress about his plans for the census. Paragraph (1) instructs him to submit, at least three years before the census date, a report containing his "determination of the subjects proposed to be included, and the types of information to be compiled," in the census. Paragraph (2) then tells him to submit, at least two years before the census date, a report containing his "determination of the questions proposed to be included" in the census. Paragraph (3) provides:

"[A]fter submission of a report under paragraph (1) or (2) of this subsection and before the appropriate census date, if the Secretary finds new circumstances exist which necessitate that the subjects, types of information, or questions contained in reports so submitted be modified, [he shall submit] a report containing the Secretary's determination of the subjects, types of information, or questions as proposed to be modified."

The Secretary timely submitted his paragraph (1) report in March 2017. It did not mention citizenship. In December 2017, he received DOJ's formal request. Three months later, in March 2018, he timely submitted his paragraph (2) report. It did propose asking a question about citizenship.

The District Court held that the Secretary's failure to mention citizenship in his March 2017 report violated § 141(f)(1) and provided an independent basis to set aside his action. Assuming without deciding that the Secretary's compliance with the reporting requirement is for courts-rather than Congress-to police, we disagree. The Secretary's March 2018 report satisfied the requirements of paragraph (3): By informing Congress that he proposed to include a citizenship question, the Secretary necessarily also informed Congress that he proposed to modify the original list of subjects that he submitted in the March 2017 report. Nothing *2573in § 141(f) suggests that the same report cannot simultaneously fulfill the requirements of paragraphs (2) and (3). And to the extent paragraph (3) requires the Secretary to explain his finding of new circumstances, he did so in his March 2018 memo, which described DOJ's intervening request.

In any event, even if we agreed with the District Court that the Secretary technically violated § 141(f) by submitting a paragraph (2) report that doubled as a paragraph (3) report, the error would surely be harmless in these circumstances, where the Secretary nonetheless fully informed Congress of, and explained, his decision. See 5 U. S. C. § 706 (in reviewing agency action, "due account shall be taken of the rule of prejudicial error").

V

We now consider the District Court's determination that the Secretary's decision must be set aside because it rested on a pretextual basis, which the Government conceded below would warrant a remand to the agency.

We start with settled propositions. First, in order to permit meaningful judicial review, an agency must "disclose the basis" of its action. Burlington Truck Lines, Inc. v. United States , 371 U.S. 156, 167-169, 83 S.Ct. 239, 9 L.Ed.2d 207 (1962) (internal quotation marks omitted); see also SEC v. Chenery Corp. , 318 U.S. 80, 94, 63 S.Ct. 454, 87 L.Ed. 626 (1943) ("[T]he orderly functioning of the process of review requires that the grounds upon which the administrative agency acted be clearly disclosed and adequately sustained.").

Second, in reviewing agency action, a court is ordinarily limited to evaluating the agency's contemporaneous explanation in light of the existing administrative record. Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Council, Inc. , 435 U.S. 519, 549, 98 S.Ct. 1197, 55 L.Ed.2d 460 (1978) ; Camp v. Pitts , 411 U.S. 138, 142-143, 93 S.Ct. 1241, 36 L.Ed.2d 106 (1973) (per curiam ). That principle reflects the recognition that further judicial inquiry into "executive motivation" represents "a substantial intrusion" into the workings of another branch of Government and should normally be avoided. Arlington Heights v. Metropolitan Housing Development Corp. , 429 U.S. 252, 268, n. 18, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977) ; see Overton Park , 401 U.S. at 420, 91 S.Ct. 814.

Third, a court may not reject an agency's stated reasons for acting simply because the agency might also have had other unstated reasons. See Jagers v. Federal Crop Ins. Corp. , 758 F.3d 1179, 1185-1186 (CA10 2014) (rejecting argument that "the agency's subjective desire to reach a particular result must necessarily invalidate the result, regardless of the objective evidence supporting the agency's conclusion"). Relatedly, a court may not set aside an agency's policymaking decision solely because it might have been influenced by political considerations or prompted by an Administration's priorities. Agency policymaking is not a "rarified technocratic process, unaffected by political considerations or the presence of Presidential power." Sierra Club v. Costle , 657 F.2d 298, 408 (CADC 1981). Such decisions are routinely informed by unstated considerations of politics, the legislative process, public relations, interest group relations, foreign relations, and national security concerns (among others).

Finally, we have recognized a narrow exception to the general rule against inquiring into "the mental processes of administrative decisionmakers." Overton Park , 401 U.S. at 420, 91 S.Ct. 814. On a *2574"strong showing of bad faith or improper behavior," such an inquiry may be warranted and may justify extra-record discovery. Ibid.

The District Court invoked that exception in ordering extra-record discovery here. Although that order was premature, we think it was ultimately justified in light of the expanded administrative record. Recall that shortly after this litigation began, the Secretary, prodded by DOJ, filed a supplemental memo that added new, pertinent information to the administrative record. The memo disclosed that the Secretary had been considering the citizenship question for some time and that Commerce had inquired whether DOJ would formally request reinstatement of the question. That supplemental memo prompted respondents to move for both completion of the administrative record and extra-record discovery. The District Court granted both requests at the same hearing, agreeing with respondents that the Government had submitted an incomplete administrative record and that the existing evidence supported a prima facie showing that the VRA rationale was pretextual.

The Government did not challenge the court's conclusion that the administrative record was incomplete, and the parties stipulated to the inclusion of more than 12,000 pages of internal deliberative materials as part of the administrative record, materials that the court later held were sufficient on their own to demonstrate pretext. The Government did, however, challenge the District Court's order authorizing extra-record discovery, as well as the court's later orders compelling depositions of the Secretary and of the Acting Assistant Attorney General for DOJ's Civil Rights Division.

We agree with the Government that the District Court should not have ordered extra-record discovery when it did. At that time, the most that was warranted was the order to complete the administrative record. But the new material that the parties stipulated should have been part of the administrative record-which showed, among other things, that the VRA played an insignificant role in the decisionmaking process-largely justified such extra-record discovery as occurred (which did not include the deposition of the Secretary himself). We accordingly review the District Court's ruling on pretext in light of all the evidence in the record before the court, including the extra-record discovery.

That evidence showed that the Secretary was determined to reinstate a citizenship question from the time he entered office; instructed his staff to make it happen; waited while Commerce officials explored whether another agency would request census-based citizenship data; subsequently contacted the Attorney General himself to ask if DOJ would make the request; and adopted the Voting Rights Act rationale late in the process. In the District Court's view, this evidence established that the Secretary had made up his mind to reinstate a citizenship question "well before" receiving DOJ's request, and did so for reasons unknown but unrelated to the VRA. 351 F.Supp.3d at 660.

The Government, on the other hand, contends that there was nothing objectionable or even surprising in this. And we agree-to a point. It is hardly improper for an agency head to come into office with policy preferences and ideas, discuss them with affected parties, sound out other agencies for support, and work with staff attorneys to substantiate the legal basis for a preferred policy. The record here reflects the sometimes involved nature of Executive Branch decisionmaking, but no *2575particular step in the process stands out as inappropriate or defective.

And yet, viewing the evidence as a whole, we share the District Court's conviction that the decision to reinstate a citizenship question cannot be adequately explained in terms of DOJ's request for improved citizenship data to better enforce the VRA. Several points, considered together, reveal a significant mismatch between the decision the Secretary made and the rationale he provided.

The record shows that the Secretary began taking steps to reinstate a citizenship question about a week into his tenure, but it contains no hint that he was considering VRA enforcement in connection with that project. The Secretary's Director of Policy did not know why the Secretary wished to reinstate the question, but saw it as his task to "find the best rationale." Id., at 551. The Director initially attempted to elicit requests for citizenship data from the Department of Homeland Security and DOJ's Executive Office for Immigration Review, neither of which is responsible for enforcing the VRA. After those attempts failed, he asked Commerce staff to look into whether the Secretary could reinstate the question without receiving a request from another agency. The possibility that DOJ's Civil Rights Division might be willing to request citizenship data for VRA enforcement purposes was proposed by Commerce staff along the way and eventually pursued.

Even so, it was not until the Secretary contacted the Attorney General directly that DOJ's Civil Rights Division expressed interest in acquiring census-based citizenship data to better enforce the VRA. And even then, the record suggests that DOJ's interest was directed more to helping the Commerce Department than to securing the data. The December 2017 letter from DOJ drew heavily on contributions from Commerce staff and advisors. Their influence may explain why the letter went beyond a simple entreaty for better citizenship data-what one might expect of a typical request from another agency-to a specific request that Commerce collect the data by means of reinstating a citizenship question on the census. Finally, after sending the letter, DOJ declined the Census Bureau's offer to discuss alternative ways to meet DOJ's stated need for improved citizenship data, further suggesting a lack of interest on DOJ's part.

Altogether, the evidence tells a story that does not match the explanation the Secretary gave for his decision. In the Secretary's telling, Commerce was simply acting on a routine data request from another agency. Yet the materials before us indicate that Commerce went to great lengths to elicit the request from DOJ (or any other willing agency). And unlike a typical case in which an agency may have both stated and unstated reasons for a decision, here the VRA enforcement rationale-the sole stated reason-seems to have been contrived.

We are presented, in other words, with an explanation for agency action that is incongruent with what the record reveals about the agency's priorities and decisionmaking process. It is rare to review a record as extensive as the one before us when evaluating informal agency action-and it should be. But having done so for the sufficient reasons we have explained, we cannot ignore the disconnect between the decision made and the explanation given. Our review is deferential, but we are "not required to exhibit a naiveté from which ordinary citizens are free." United States v. Stanchich , 550 F.2d 1294, 1300 (CA2 1977) (Friendly, J.). The reasoned explanation requirement of administrative law, after all, is meant to ensure that agencies offer genuine justifications for important *2576decisions, reasons that can be scrutinized by courts and the interested public. Accepting contrived reasons would defeat the purpose of the enterprise. If judicial review is to be more than an empty ritual, it must demand something better than the explanation offered for the action taken in this case.

In these unusual circumstances, the District Court was warranted in remanding to the agency, and we affirm that disposition. See Florida Power & Light Co. v. Lorion , 470 U.S. 729, 744, 105 S.Ct. 1598, 84 L.Ed.2d 643 (1985). We do not hold that the agency decision here was substantively invalid. But agencies must pursue their goals reasonably. Reasoned decisionmaking under the Administrative Procedure Act calls for an explanation for agency action. What was provided here was more of a distraction.

* * *

The judgment of the United States District Court for the Southern District of New York is affirmed in part and reversed in part, and the case is remanded for further proceedings consistent with this opinion.

It is so ordered.

Justice THOMAS, with whom Justice GORSUCH and Justice KAVANAUGH join, concurring in part and dissenting in part.

In March 2018, the Secretary of Commerce exercised his broad discretion over the administration of the decennial census to resume a nearly unbroken practice of asking a question relating to citizenship. Our only role in this case is to decide whether the Secretary complied with the law and gave a reasoned explanation for his decision. The Court correctly answers these questions in the affirmative. Ante , at 2566 - 2573. That ought to end our inquiry.

The Court, however, goes further. For the first time ever, the Court invalidates an agency action solely because it questions the sincerity of the agency's otherwise adequate rationale. Echoing the din of suspicion and distrust that seems to typify modern discourse, the Court declares the Secretary's memorandum "pretextual" because, "viewing the evidence as a whole," his explanation that including a citizenship question on the census would help enforce the Voting Rights Act (VRA) "seems to have been contrived." Ante , at 2572 - 2573, 2574 - 2575, 2575 - 2576. The Court does not hold that the Secretary merely had additional , unstated reasons for reinstating the citizenship question. Rather, it holds that the Secretary's stated rationale did not factor at all into his decision.

The Court's holding reflects an unprecedented departure from our deferential review of discretionary agency decisions. And, if taken seriously as a rule of decision, this holding would transform administrative law. It is not difficult for political opponents of executive actions to generate controversy with accusations of pretext, deceit, and illicit motives. Significant policy decisions are regularly criticized as products of partisan influence, interest-group pressure, corruption, and animus. Crediting these accusations on evidence as thin as the evidence here could lead judicial review of administrative proceedings to devolve into an endless morass of discovery and policy disputes not contemplated by the Administrative Procedure Act (APA).

Unable to identify any legal problem with the Secretary's reasoning, the Court imputes one by concluding that he must not be telling the truth. The Court therefore upholds the decision of the District Court-which, in turn, was transparently based on the application of an administration-specific standard. App. to Pet. for Cert. 527a (crediting respondents' allegations *2577that "the current Department of Justice has shown little interest in enforcing the" VRA (emphasis added)).

The law requires a more impartial approach. Even assuming we are authorized to engage in the review undertaken by the Court-which is far from clear-we have often stated that courts reviewing agency action owe the Executive a "presumption of regularity." Citizens to Preserve Overton Park, Inc. v. Volpe , 401 U.S. 402, 415, 91 S.Ct. 814, 28 L.Ed.2d 136 (1971). The Court pays only lipservice to this principle. But, the evidence falls far short of supporting its decision. The Court, I fear, will come to regret inventing the principles it uses to achieve today's result. I respectfully dissent from Part V of the opinion of the Court.1

I

As the Court explains, federal law directs the Secretary of Commerce to "take a decennial census." 13 U. S. C. § 141(a) ; see U. S. Const., Art. I, § 2, cl. 3 ; Amdt. XIV, § 2; ante , at 2561 - 2562. The discretion afforded the Secretary is extremely broad. Subject only to constitutional limitations and a handful of inapposite statutory requirements, the Secretary is expressly authorized to "determine the inquiries" on the census questionnaire and to conduct the census "in such form and content as he may determine." §§ 5, 141(a) ; see ante , at 2567 - 2569, 2571 - 2573.2 Prior census questionnaires have included questions ranging from sex, age, and race to commute, education, and radio ownership. And between 1820 and 2010, every decennial census questionnaire but one asked some segment of the population a question related to citizenship. The 2010 census was the first since 1840 that did not include any such question.

In March 2018, the Secretary issued a memorandum reinstating a citizenship question on the 2020 census. He explained that the Department of Justice (DOJ) had formally requested reinstatement of the question because the data obtained would help enforce § 2 of the VRA. He further explained that the question had been well tested in light of its extensive previous use, that he had consulted with the Census Bureau on the proposal, and that his final decision incorporated feedback from the Bureau. He recognized that staff at the Bureau believed that better data could be obtained through modeling and reliance on existing records, but he disagreed with that assessment, explaining that the data was inconclusive and that he thought it preferable to ask the question directly of the entire population. Respondents brought suit, seeking judicial review of the Secretary's decision under the APA, 5 U. S. C. § 706.

II

As relevant here, the APA requires courts to "hold unlawful and set aside" agency action that is "arbitrary, capricious, an abuse of discretion, or otherwise not in *2578accordance with law." § 706(2)(A). We have emphasized that "[r]eview under the arbitrary and capricious standard is deferential." National Assn. of Home Builders v. Defenders of Wildlife , 551 U.S. 644, 658, 127 S.Ct. 2518, 168 L.Ed.2d 467 (2007) ; see Glickman v. Wileman Brothers & Elliott, Inc. , 521 U.S. 457, 466, n. 8, 117 S.Ct. 2130, 138 L.Ed.2d 585 (1997). It requires the reviewing court to determine whether the agency " 'examine[d] the relevant data and articulate[d] a satisfactory explanation for its action.' " FCC v. Fox Television Stations, Inc. , 556 U.S. 502, 513, 129 S.Ct. 1800, 173 L.Ed.2d 738 (2009). We have described this as a " 'narrow' standard of review" under which the reviewing court cannot " 'substitute its judgment for that of the agency,' and should 'uphold a decision of less than ideal clarity if the agency's path may reasonably be discerned.' " Id. , at 513-514, 129 S.Ct. 1800 (citation omitted); accord, Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co. , 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983).3

Part IV-B of the opinion of the Court correctly applies this standard to conclude that the Secretary's decision survives ordinary arbitrary-and-capricious review. That holding should end our inquiry.

But the opinion continues. Acknowledging that "no particular step" in the proceedings here "stands out as inappropriate or defective," even after reviewing "all the evidence in the record ..., including the extra-record discovery," ante , at 2574, the Court nevertheless agrees with the District Court that the Secretary's rationale for reinstating the citizenship question was "pretextual-that is, that the real reason for his decision was something other than the sole reason he put forward in his memorandum, namely enhancement of DOJ's VRA enforcement efforts." 351 F.Supp.3d 502, 660 (SDNY 2019) ; see ante , at 2575 - 2576. According to the Court, something just "seems" wrong. Ibid.

This conclusion is extraordinary. The Court engages in an unauthorized inquiry into evidence not properly before us to reach an unsupported conclusion. Moreover, each step of the inquiry offends the presumption of regularity we owe the Executive. The judgment of the District Court should be reversed.

A

Section 706(2) of the APA contemplates review of the administrative "record" to determine whether an agency's "action, findings, and conclusions" satisfy six specified standards. See §§ 706(2)(A)-(F). None instructs the Court to inquire into pretext. Consistent with this statutory text, we have held that a court is "ordinarily limited to evaluating the agency's contemporaneous explanation in light of the existing administrative record." Ante , at 2573 (citing Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Council, Inc. , 435 U.S. 519, 549, 98 S.Ct. 1197, 55 L.Ed.2d 460 (1978) ); see SEC v. Chenery Corp. , 318 U.S. 80, 87, 63 S.Ct. 454, 87 L.Ed. 626 (1943) ("The grounds upon which an administrative order must be judged are those upon which the record discloses that its action was based"). If an agency's stated findings and conclusions withstand scrutiny, the APA does not permit a court to set aside the decision solely because the agency had "other unstated *2579reasons" for its decision, such as "political considerations" or the "Administration's priorities." Ante , at 2573 - 2574.

Unsurprisingly, then, this Court has never held an agency decision arbitrary and capricious on the ground that its supporting rationale was "pretextual." Nor has it previously suggested that this was even a possibility. Under "settled propositions" of administrative law, ante , at 2572 - 2573, pretext is virtually never an appropriate or relevant inquiry for a reviewing court to undertake.

Respondents conceptualize pretext as a subset of "arbitrary and capricious" review. It is far from clear that they are correct. But even if they were, an agency action is not arbitrary or capricious merely because the decisionmaker has other, unstated reasons for the decision. Ante , at 2573 - 2574. Nor is an agency action arbitrary and capricious merely because the decisionmaker was "inclined" to accomplish it before confirming that the law and facts supported that inclination. In re Dept. of Commerce , 586 U. S. ----, ----, 139 S.Ct. 16, 17, 202 L.Ed.2d 306 (2018) (GORSUCH, J., concurring in part and dissenting in part).

Accordingly, even under respondents' approach, a showing of pretext could render an agency action arbitrary and capricious only in the infinitesimally small number of cases in which the administrative record establishes that an agency's stated rationale did not factor at all into the decision, thereby depriving the action of an adequate supporting rationale.4 This showing is extremely difficult to make because the administrative record will rarely, if ever, contain evidence sufficient to show that an agency's stated rationale did not actually factor into its decision. And we have stated that a "strong showing of bad faith or improper behavior" is necessary to venture beyond the agency's "administrative findings" and inquire into "the mental processes of administrative decisionmakers." Overton Park , 401 U.S. at 420, 91 S.Ct. 814.5 We have never before found Overton Park 's exception satisfied, much less invalidated an agency action based on "pretext."

Undergirding our arbitrary-and-capricious analysis is our longstanding precedent affording the Executive a "presumption of regularity." Id., at 415, 91 S.Ct. 814 ; see United States v. Chemical Foundation, Inc. , 272 U.S. 1, 14-15, 47 S.Ct. 1, 71 L.Ed. 131 (1926). This presumption reflects respect for a coordinate branch of government whose officers not only take *2580an oath to support the Constitution, as we do, Art. VI, but also are charged with "faithfully execut[ing]" our laws, Art. II, § 3. See United States v. Morgan , 313 U.S. 409, 422, 61 S.Ct. 999, 85 L.Ed. 1429 (1941) (presumption of regularity ensures that the "integrity of the administrative process" is appropriately respected). In practice, then, we give the benefit of the doubt to the agency.

B

The Court errs at the outset by proceeding beyond the administrative record to evaluate pretext. Respondents have not made a "strong showing of bad faith or improper behavior." Overton Park , supra , at 420, 91 S.Ct. 814.

The District Court's initial order granting extra-record discovery relied on four categories of evidence:

"evidence that [the Secretary] was predisposed to reinstate the citizenship question when he took office; that the [DOJ] hadn't expressed a desire for more detailed citizenship data until the Secretary solicited its view; that he overruled the objections of his agency's career staff; and that he declined to order more testing of the question given its long history." Dept. of Commerce , 586 U. S., at ----, 139 S.Ct., at 18.

None of this comes close to showing bad faith or improper behavior. Indeed, there is nothing even "unusual about a new cabinet secretary coming to office inclined to favor a different policy direction, soliciting support from other agencies to bolster his views, disagreeing with staff, or cutting through red tape." Ibid . Today all Members of the Court who reach the question agree that the District Court abused its discretion in ordering extra-record discovery based on this evidence. Ante , at 2574 ("We agree with the Government that the District Court should not have ordered extra-record discovery when it did").

Nevertheless, the Court excuses the error because, in its view, "the new material that the parties [later] stipulated should have been part of the administrative record ... largely justified such extra-record discovery as occurred." Ibid. Given the requirement that respondents make a "strong showing" of bad faith, one would expect the Court to identify which "new material" supported such a showing. It does not. Nor does the Court square its suggestion that some of the extra-record discovery was not "justified" with its consideration of "all ... the extra-record discovery." Ante , at 2574 - 2575. Regardless, I assume that the Court has in mind the administrative-record materials that the District Court would later rely on to establish pretext:

"evidence that [the Secretary] had made the decision to add the citizenship question well before DOJ requested its addition in December 2017; the absence of any mention, at all , of VRA enforcement in the discussions of adding the question that preceded the [DOJ] Letter; unsuccessful attempts by Commerce Department staff to shop around for a request by another agency regarding citizenship data; and [the Secretary's] personal outreach to Attorney General Sessions, followed by the [DOJ] Letter; not to mention the conspicuous procedural irregularities that accompanied the decision to add the question." 351 F.Supp.3d at 661 (citations omitted).

This evidence fails to make a strong showing of bad faith or improper behavior. Taken together, it proves at most that the Secretary was predisposed to add a citizenship question to the census and took steps to achieve that end before settling on the VRA rationale he included in his memorandum. Perhaps he had reasons for adding *2581the citizenship question other than the VRA, but by the Court's own telling, that does not amount to evidence of bad faith or improper behavior. Ante , at 2573 - 2574; see Dept. of Commerce , supra , at 17.

The Court thus errs in relying on materials outside the record to support its holding. And the Court does not claim that the evidence in the administrative record alone would prove that the March 2018 memorandum was a pretext. Given the presumption of regularity, the evidence discussed above falls far short of establishing that the VRA rationale did not factor at all into the Secretary's decision.

C

Even if it were appropriate for the Court to rely on evidence outside the administrative record, that evidence still fails to establish pretext. None of the evidence cited by the Court or the District Court comes close to showing that the Secretary's stated rationale-that adding a citizenship question to the 2020 census questionnaire would "provide ... data that are not currently available" and "permit more effective enforcement of the [VRA]," App. to Pet. for Cert. 548a-did not factor at all into his decision.

Once again, the evidence cited by the Court suggests at most that the Secretary had "other unstated reasons" for reinstating the citizenship question. Ante , at 2573 - 2574. For example, the Court states that the Secretary's Director of Policy "initially attempted to elicit requests for citizenship data from the Department of Homeland Security and DOJ's Executive Office for Immigration Review." Ante , at 2575. But this hardly shows pretext. It simply suggests that the Director believed that citizenship information could be useful in tackling problems related to national security and illegal immigration-a view that would also explain why the Secretary might not have been "considering VRA enforcement" early on. Ibid. ; see also American Community Survey, Why We Ask: Place of Birth, Citizenship and Year of Entry (2016) (explaining that inquiries about "place of birth, citizenship, and year of entry" provide statistics that are "essential for agencies and policy makers setting and evaluating immigration policies and laws, understanding how different immigrant groups are assimilated, and monitoring against discrimination"), https://www2.census.gov/programs-surveys/acs/about/qbyqfact/2016/Citizenship.pdf (as last visited June 25, 2019).

The Court emphasizes that the VRA rationale for the citizenship question originated in the Department of Commerce, and suggests that DOJ officials unthinkingly fell in line after the Attorney General was looped into the process. See ante , at 2575. But the Court ignores that the letter was drafted by the then-Acting Assistant Attorney General for Civil Rights and reviewed by five other DOJ attorneys, including the Chief of the DOJ's Voting Section. 351 F.Supp.3d at 554-556. Given the DOJ's multilayer review process and its explanation for requesting citizenship data, the Court's suggestion that the DOJ's letter was inadequately vetted or improperly "influence[d]" by the Department of Commerce is entirely unsupported. Ante , at 2575. In any event, none of this suggests, much less proves, that the Secretary harbored an unstated belief that adding the citizenship question would not help enforce the VRA, or that the VRA rationale otherwise did not factor at all into his decision. It simply suggests that a number of executive officials agreed that adding a citizenship question would support VRA enforcement.

The Court's other evidence is even further afield. The Court thinks it telling that the DOJ's letter included "a specific request *2582that Commerce collect the [citizenship] data by means of reinstating a citizenship question on the census," rather than a more open-ended "entreaty for better citizenship data." Ibid. I do not understand how the specificity of the DOJ's letter bears on whether the Secretary's rationale was pretextual-particularly since the letter specifically explained why "census questionnaire data regarding citizenship, if available, would be more appropriate for use in redistricting and in [VRA] litigation" than existing data. App. to Pet. for Cert. 568a; see id. , at 567a-568a. Unless the Court is now suggesting that agency correspondence must comply with the Court's subjective, unsupported view of what "might" constitute a "typical request from another agency," ante , at 2575, the specificity of the DOJ's letter is irrelevant. The Court also points to the DOJ's decision not to meet with the Census Bureau "to discuss alternative ways to meet DOJ's stated need for improved citizenship data." Ibid. But the Court does not explain how the DOJ's refusal bears on the Secretary's rationale. Besides, it is easy to understand why DOJ officials would not be interested in meeting with the Census Bureau. The meeting would have been with career employees whose acknowledged purpose was to talk the DOJ out of its request. See 351 F.Supp.3d at 557. Having already considered the issue and explained the rationale behind the request, it seems at least plausible that the DOJ officials believed such a meeting would be unproductive.

In short, the evidence cited by the Court establishes, at most, that leadership at both the Department of Commerce and the DOJ believed it important-for a variety of reasons-to include a citizenship question on the census.

The Court also fails to give credit where it is due. The Secretary initiated this process inclined to favor what he called "Option B"-that is, simply "add[ing] a citizenship question to the decennial census." App. to Pet. for Cert. 552a. But the Census Bureau favored "Option C"-relying solely on "administrative records" to supply the information needed by the DOJ. Id. , at 554a. The Secretary considered this view and found it a "potentially appealing solution," ibid. , but concluded that it had shortcomings. Rather than revert to his original inclination, however, he "asked the Census Bureau to develop a fourth alternative, Option D, which would combine Options B and C." Id. , at 555a. And he settled on that solution. Whatever one thinks of the Secretary's choice, his willingness to change his mind in light of the Bureau's feedback belies the idea that his rationale or decisionmaking process was a pretext.

The District Court's lengthy opinion pointed to other facts that, in its view, supported a finding of pretext. 351 F.Supp.3d at 567-572, 660-664 (discussing the statements, e-mails, acts, and omissions of numerous people involved in the process). I do not deny that a judge predisposed to distrust the Secretary or the administration could arrange those facts on a corkboard and-with a jar of pins and a spool of string-create an eye-catching conspiracy web. Cf. id. , at 662 (inferring "from the various ways in which [the Secretary] and his aides acted like people with something to hide that they did have something to hide"). But the Court does not rely on this evidence, and rightly so: It casts no doubt on whether the Secretary's stated rationale factored into his decision. The evidence suggests, at most, that the Secretary had multiple reasons for wanting to include the citizenship question on the census.

Finally, if there could be any doubt about this conclusion, the presumption of *2583regularity resolves it. Where there are equally plausible views of the evidence, one of which involves attributing bad faith to an officer of a coordinate branch of Government, the presumption compels giving the benefit of the doubt to that officer.

III

The Court's erroneous decision in this case is bad enough, as it unjustifiably interferes with the 2020 census. But the implications of today's decision are broader. With today's decision, the Court has opened a Pandora's box of pretext-based challenges in administrative law.

Today's decision marks the first time the Court has ever invalidated an agency action as "pretextual." Having taken that step, one thing is certain: This will not be the last time it is asked to do so. Virtually every significant agency action is vulnerable to the kinds of allegations the Court credits today. These decisions regularly involve coordination with numerous stakeholders and agencies, involvement at the highest levels of the Executive Branch, opposition from reluctant agency staff, and-perhaps most importantly-persons who stand to gain from the action's demise. Opponents of future executive actions can be expected to make full use of the Court's new approach.

The 2015 "Open Internet Order" provides a case in point. In 2015, the Federal Communications Commission (FCC) adopted a controversial order reclassifying broadband Internet access service as a "telecommunications service" subject to regulation under Title II of the Communications Act. See In re Protecting and Promoting the Open Internet , 30 FCC Rcd. 5601, 5618 (2015). According to a dissenting Commissioner, the FCC "flip-flopp[ed]" on its previous policy not because of a change in facts or legal understanding, but based on "one reason and one reason alone. President Obama told us to do so." Id., at 5921 (statement of Comm'r Pai). His view was supported by a 2016 congressional Report in which Republican Senate staff concluded that "the FCC bent to the political pressure of the White House" and "failed to live up to standards of transparency." Majority Staff Report, Senate Committee on Homeland Security and Governmental Affairs, Regulating the Internet: How the White House Bowled Over FCC Independence, 114th Cong., 1st Sess., 29 (Comm. Print 2016). The Report cited evidence strikingly similar to that relied upon by the Court here-including agency-initiated "meetings with certain outside groups to support" the new result, id. , at 3; "apparen[t] ... concern from the career staff that there was insufficient notice to the public and affected stakeholders," id., at 4; and "regula[r] communicatio[n]" between the FCC Chairman and "presidential advisors," id. , at 25.

Under the malleable standard applied by the Court today, a serious case could be made that the Open Internet Order should have been invalidated as "pretextual," regardless of whether any "particular step in the process stands out as inappropriate or defective." Ante , at 2575. It is enough, according to the Court, that a judge believes that the ultimate rationale "seems to have been contrived" when the evidence is considered "as a whole." Ante , at 2574, 2575 - 2576.

Now that the Court has opened up this avenue of attack, opponents of executive actions have strong incentives to craft narratives that would derail them. Moreover, even if the effort to invalidate the action is ultimately unsuccessful, the Court's decision enables partisans to use the courts to harangue executive officers through depositions, discovery, delay, and distraction. The Court's decision could even implicate separation-of-powers concerns insofar as it *2584enables judicial interference with the enforcement of the laws.

In short, today's decision is a departure from traditional principles of administrative law. Hopefully it comes to be understood as an aberration-a ticket good for this day and this train only.

* * *

Because the Secretary's decision to reinstate a citizenship question on the 2020 census was legally sound and a reasoned exercise of his broad discretion, I respectfully dissent from Part V of the opinion of the Court.

Justice BREYER, with whom Justice GINSBURG, Justice SOTOMAYOR, and Justice KAGAN join, concurring in part and dissenting in part.

I join Parts I, II, IV-A, and V of the Court's opinion (except as otherwise indicated in this opinion). I dissent, however, from the conclusion the Court reaches in Part IV-B. To be more specific, I agree with the Court that the Secretary of Commerce provided a pretextual reason for placing a question about citizenship on the short-form census questionnaire and that a remand to the agency is appropriate on that ground. But I write separately because I also believe that the Secretary's decision to add the citizenship question was arbitrary and capricious and therefore violated the Administrative Procedure Act (APA).

There is no serious dispute that adding a citizenship question would diminish the accuracy of the enumeration of the population-the sole constitutional function of the census and a task of great practical importance. The record demonstrates that the question would likely cause a disproportionate number of noncitizens and Hispanics to go uncounted in the upcoming census. That, in turn, would create a risk that some States would wrongfully lose a congressional representative and funding for a host of federal programs. And, the Secretary was told, the adverse consequences would fall most heavily on minority communities. The Secretary decided to ask the question anyway, citing a need for more accurate citizenship data. But the evidence indicated that asking the question would produce citizenship data that is less accurate, not more. And the reason the Secretary gave for needing better citizenship data in the first place-to help enforce the Voting Rights Act of 1965-was not convincing.

In short, the Secretary's decision to add a citizenship question created a severe risk of harmful consequences, yet he did not adequately consider whether the question was necessary or whether it was an appropriate means of achieving his stated goal. The Secretary thus failed to "articulate a satisfactory explanation" for his decision, "failed to consider ... important aspect[s] of the problem," and "offered an explanation for [his] decision that runs counter to the evidence," all in violation of the APA. Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co. , 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983). These failures, in my view, risked undermining public confidence in the integrity of our democratic system itself. I would therefore hold that the Secretary's decision-whether pretextual or not-was arbitrary, capricious, and an abuse of discretion.

I

A

Three sets of laws determine the legal outcome of this case. First, the Constitution requires an "actual Enumeration" of the "whole number of persons in each State" every 10 years. Art. I, § 2, cl. 3 ; Amdt. 14, § 2. It does so in order to *2585"provide a basis for apportioning representatives among the states in the Congress." Baldrige v. Shapiro , 455 U.S. 345, 353, 102 S.Ct. 1103, 71 L.Ed.2d 199 (1982) ; see also Art. I, § 2, cl. 3. The inclusion of this provision in the Constitution itself underscores the importance of conducting an accurate census. See Utah v. Evans , 536 U.S. 452, 478, 122 S.Ct. 2191, 153 L.Ed.2d 453 (2002) (recognizing "a strong constitutional interest in [the] accuracy" of the enumeration).

Second, the Census Act contains two directives that constrain the Secretary's ability to add questions to the census. Section 195 says that the Secretary "shall, if he considers it feasible," authorize the use of statistical "sampling" in collecting demographic information. That means the Secretary must, if feasible, obtain demographic information through a survey sent to a sample of households, rather than through the short-form census questionnaire to which every household must respond. The other relevant provision, § 6(c), says that "[t ]o the maximum extent possible and consistent with the kind, timeliness, quality and scope of the statistics required, the Secretary shall acquire and use information available" from administrative sources "instead of conducting direct inquiries." (Emphasis added.) These provisions, taken together, reflect a congressional preference for keeping the short form short, so that it does not burden recipients and thereby discourage them from responding.

Third, the APA prohibits administrative agencies from making choices that are "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law." 5 U. S. C. § 706(2)(A). We have said that courts, in applying this provision, must decide "whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment." Citizens to Preserve Overton Park, Inc. v. Volpe , 401 U.S. 402, 416, 91 S.Ct. 814, 28 L.Ed.2d 136 (1971). The agency must have "examine[d] the relevant data and articulate[d] a satisfactory explanation for its action[,] including a 'rational connection between the facts found and the choice made.' " State Farm , 463 U.S. at 43, 103 S.Ct. 2856. An agency ordinarily fails to meet this standard if it has "failed to consider an important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency, or is so implausible that it could not be ascribed to a difference in view or the product of agency expertise." Ibid.

Courts do not apply these principles of administrative law mechanically. Rather, they take into account, for example, the nature and importance of the particular decision, the relevance and importance of missing information, and the inadequacies of a particular explanation in light of their importance. The Federal Government makes tens of thousands, perhaps millions, of administrative decisions each year. And courts would be wrong to expect or insist upon administrative perfection. But here, the Enumeration Clause, the Census Act, and the nature of the risks created by the agency's decision all make clear that the decision before us is highly important to the proper functioning of our democratic system. It is therefore particularly important that courts here not overlook an agency's (1) failure to consider serious risks of harm, (2) failure to explain its refusal to minimize those risks, or (3) failure to link its conclusion to available evidence. My view, like that of the District Court, is that the agency here failed on all three counts.

B

A brief history of how the census has worked over the years will help the reader understand some of the shortcomings of *2586the Secretary's decisionmaking process. The Framers wrote into the Constitution a mandate to conduct an "actual Enumeration" of the population every 10 years. Art. I, § 2, cl. 3. They did so for good reason. The purpose of the census is to "provide a basis for apportioning representatives among the states in the Congress," Baldrige , 455 U.S. at 353, 102 S.Ct. 1103, ensuring that "comparative state political power in the House ... reflect[s] comparative population," Evans , 536 U.S. at 477, 122 S.Ct. 2191. The Framers required an actual count of every resident to "limit political chicanery" and to prevent the census count from being "skewed for political ... purposes." Id. , at 500, 122 S.Ct. 2191 (THOMAS, J., concurring in part and dissenting in part).

Throughout most of the Nation's history, the Federal Government used enumerators, often trained census takers, to conduct the census by going door to door. The enumerators would ask a host of questions, including place of birth, citizenship, and others. But after the 1950 census, the Bureau began to change its approach. Post-census studies revealed that the census had failed to count more than 5 million people and that the undercount disproportionately affected members of minority groups. See M. Anderson, The American Census: A Social History 201-202 (1988); Brief for Historians and Social Scientists as Amici Curiae 15. Studies showed that statistical sampling would produce higher quality data. Anderson, American Census, at 201.

Beginning with the 1960 census, the Bureau consequently divided its questioning into a short form and a long form. The short form contained a list of questions-a short list-that the census would ask of every household. That list included basic demographic questions like sex, age, race, and marital status. The short form did not include, and has never included, a question about citizenship. See ibid. ; Dept. of Commerce, U. S. Census Bureau, Measuring America: The Decennial Censuses From 1790 to 2000, p. 128 (2002). By way of contrast, the long form set forth a host of questions that would be asked of only a sample of households. In 1960, the long form was sent to one in every four households; in subsequent years, it was sent to approximately one in every six. See 351 F.Supp.3d 502, 520 (SDNY 2019). And it was more recently replaced by the American Community Survey (ACS), which is sent to approximately 1 in 38 households each year. The long form (and now the ACS) has often included a question about citizenship.

In 1970, the Census Bureau made another important change to the census. It significantly reduced its reliance upon in-person enumerators. See Anderson, supra , at 206. Instead, it sent nearly all households a questionnaire by mail. Most households received the short form, and a small sample received the long form. Instructions on the form told each household to fill out the questionnaire and return it to the Census Bureau by mail. Enumerators would follow up with households that did not return the questionnaire.

To maximize accuracy and minimize cost, the Bureau tried to bring about the highest possible "self-response" rate, i.e. , to encourage as many households as possible to respond by mail. For that reason, it tried to keep the short form as short as possible. And it consistently opposed placing a citizenship question on that form. It feared that adding a question about citizenship would "inevitably jeopardize the overall accuracy of the population count," partly because of added response burden but also because, as it explained, noncitizens faced with a citizenship question would be less likely to respond due to *2587fears of "the information being used against them." Federation for Am. Immigration Reform v. Klutznick , 486 F.Supp. 564, 568 (DDC 1980).

Likely for similar reasons, Congress amended the Census Act in 1976, enacting the two statutory provisions to which I previously referred. These two provisions, 13 U. S. C. § 6(c) and § 195, together encourage the Secretary not to ask demographic questions on the short form if the information can be obtained either through the long form or through administrative records.

II

With this statutory and historical background, we can more easily consider the agency decision directly under review. That decision "reinstate[s] [a] citizenship question on the 2020 decennial census." App. to Pet. for Cert. 549a-550a (Memorandum from Wilbur L. Ross, Jr., Secretary of Commerce, to Karen Dunn Kelley, Under Secretary for Economic Affairs (Mar. 26, 2018)). The agency's decision memorandum provided one and only one reason for making that decision-namely, that the question was "necessary to provide complete and accurate data in response to" a request from the Department of Justice (DOJ). Id. , at 562a. The DOJ had requested the citizenship question for "use [in] ... determining violations of Section 2 of the Voting Rights Act." Id. , at 548a.

The decision memorandum adds that the agency had not been able to "determine definitively how inclusion of a citizenship question on the decennial census will impact responsiveness. However, even if there is some impact on responses, the value of more complete and accurate data derived from surveying the entire population outweighs such concerns." Id. , at 562a. The Secretary's decision thus rests upon a weighing of potentially adverse consequences (diminished responses and a less accurate census count) against potentially offsetting advantages (better citizenship data). In my view, however, the Secretary did not make reasonable decisions about these potential costs and benefits in light of the administrative record.

A

Consider first the Secretary's conclusion that he was "not able to determine definitively how inclusion of a citizenship question on the decennial census will impact responsiveness." Ibid. Insofar as this statement implies that adding the citizenship question is unlikely to affect "responsiveness" very much (or perhaps at all), the evidence in the record indicates the contrary.

1

The administrative record includes repeated Census Bureau statements that adding the question would produce a less accurate count because noncitizens and Hispanics would be less likely to respond to the questionnaire. See App. 105, 109-112, 158. The Census Bureau's chief scientist said specifically that adding the question would have "an adverse impact on self-response and, as a result, on the accuracy and quality of the 2020 Census." Id. , at 109. And the chief scientist backed this statement up by pointing to "[t]hree distinct analyses." Ibid.

The first analysis compared nonresponse rates for the short-form census questionnaire (which did not include a citizenship question) to nonresponse rates for the ACS (which did). Obviously, more people fail to respond to the ACS than to the short form. Yet taking into account the fact that the nonresponse rate will be greater for the ACS than for the short form, the Bureau found that the difference *2588between the two is yet greater for noncitizen households than for citizen households (by 5.1%, according to the Bureau). Id. , at 111. This led the Bureau to say that it was a "reasonable inference" that the presence of the citizenship question accounted for the difference. Ibid.

The Bureau conducted two additional studies, both analyzing data from the ACS. One study looked at response rates for particular questions on the ACS. It showed that the "no answer" rate for the citizenship question was "much greater than the comparable rates" for other census questions (for example, questions about age, sex, race, and ethnicity). Id. , at 110. And it showed that the "no answer" rate for the citizenship question was significantly higher among Hispanics. Id. , at 109-110. The last study examined "break-off" rates, i.e. , the rate at which respondents stopped answering the questionnaire upon reaching a particular question. It found that Hispanics were significantly more likely than were non-Hispanics to stop answering at the point they reached the citizenship question. Id. , at 112. Together, these two studies provided additional support for the Census Bureau's determination that the citizenship question is likely to mean disproportionately fewer responses from noncitizens and Hispanics than from others. Ibid.

Putting numbers upon these study results, the Census Bureau estimated that adding the question to the short form would lead to 630,000 additional nonresponding households. Id. , at 114. That is to say, the question would cause households covering more than 1 million additional people to decline to respond to the census. When the Bureau does not receive a response, it follows up with in-person interviews in an effort to obtain the missing information. The Bureau often interviews what it calls "proxies," such as family members and neighbors. But this followup process is subject to error; and the error rate is much greater than the error rate for self-responses. Ibid. The Bureau thus explained that lower self-response rates "degrade data quality" by increasing the risk of error and leading to hundreds of thousands of fewer correct enumerations. Id. , at 113-115. The Bureau added that its estimate was "conservative." Id. , at 115. It expected "differences between citizen and noncitizen response rates and data quality" to be "amplified" in the 2020 census "compared to historical levels." Ibid. Thus, it explained, "the decrease in self-response for citizen households in 2020 could be much greater than the 5.1 percentage points [it] observed during the 2010 Census." Id. , at 115-116. Its conclusion in light of this evidence was clear. Adding the citizenship question to the short form was "very likely to reduce the self-response rate" and thereby "har[m] the quality of the census count." Id. , at 105, 158.

The Census Bureau's analysis received support from other submissions. Several States pointed out that noncitizens and racial minorities had been undercounted in every prior census. Administrative Record 1091-1092. They also drew attention to recent surveys indicating that noncitizens had significant concerns about the confidentiality of census responses. Ibid. Former directors of the Census Bureau wrote that adding the citizenship question so late in the process "would put the accuracy of the enumeration and success of the census in all communities at grave risk." Id. , at 1057. The American Sociological Association and Census Scientific Advisory Committee echoed these warnings. See id. , at 787, 794-795. On the other hand, the Secretary received submissions by other groups that supported adding the question. See, e.g. , id. , at 1178-1179, 1206, 1276. But as far as I can tell (or as far as the *2589arguments made here and in the District Court inform the matter), none of these latter submissions significantly added to, or detracted from, the Census Bureau's submissions in respect to the question's likely impact on response rates.

2

The Secretary's decision memorandum reached a quite different conclusion from the Census Bureau. The memorandum conceded that "a lower response rate would lead to ... less accurate responses." App. to Pet. for Cert. 556a. But it concluded that neither the Census Bureau nor any stakeholders had provided "definitive, empirical support" for the proposition that the citizenship question would reduce response rates. Id. , at 554a. The memorandum relied for that conclusion upon a number of considerations, but each is contradicted by the record.

The memorandum first pointed to perceived shortcomings in the Census Bureau's analysis of nonresponse rates. It noted that response rates are generally lower overall for the long form and ACS than they are for the short form. Id. , at 552a-554a. But the Bureau explained that its analysis accounted for this consideration, see App. 111, and no one has given us reason to think the contrary. The Secretary also noted that the Bureau "was not able to isolate what percentage of [the] decline was caused by the inclusion of a citizenship question rather than some other aspect of the long form survey." App. to Pet. for Cert. 554a. But the Bureau said attributing the decline to the citizenship question was a "reasonable inference," App. 111, and again, nothing in the record contradicted the Bureau's judgment. And later analyses have borne out the Bureau's judgment that the citizenship question contributes to the decline in self-response. See, e.g. , id. , at 1002-1006, 1008 (August 2018 Census Bureau study).

The memorandum next cast doubt on the Census Bureau's analysis of the rate at which people responded to particular questions on the ACS. It noted that the "no answer" rate to the citizenship question was comparable to the "no answer" rate for other questions on the ACS, including educational attainment, income, and property insurance. App. to Pet. for Cert. 553a. But as discussed above, the Bureau found it significant that the "no answer" rate for the citizenship question was "much greater" than the "no answer" rate for the other questions that appear on the short form -that is, the form on which the citizenship question would appear. App. 110, 124. The Secretary offered no reason why the demographic variables to which he pointed provided a better point of comparison.

Finally, the memorandum relied on information provided by two outside stakeholders. The first was a study conducted by the private survey company Nielsen, in which questions about place of birth and time of arrival had not led to any appreciable decrease in the response rate. App. to Pet. for Cert. 552a. But Nielsen, which in fact urged the Secretary not to add the question, stated that its respondents (unlike census respondents) were paid to respond, and it is consequently not surprising that they did so. Administrative Record 1276. The memorandum also cited statements by former Census Bureau officials suggesting that empirical evidence about the question's potential impact on response rates was "limited." App. to Pet. for Cert. 558a-559a; see also id. , at 552a. But there was no reason to expect the former officials to provide more extensive empirical evidence as to a citizenship question when they were not privy to the internal Bureau analyses on this question. And, like Nielsen, the former *2590officials strongly urged the Secretary not to ask the question. See Administrative Record 1057.

The upshot is that the Secretary received evidence of a likely drop in census accuracy by a number somewhere in the hundreds of thousands, and he received nothing significant to the contrary. The Secretary pointed out that the Census Bureau's information was uncertain, i.e. , not "definitive." But that is not a satisfactory answer. Few public-policy-related statistical studies of risks (say, of many health or safety matters) are definitive. As the Court explained in State Farm , "[i]t is not infrequent that the available data do not settle a regulatory issue, and the agency must then exercise its judgment in moving from the facts and probabilities on the record to a policy conclusion." 463 U.S. at 52, 103 S.Ct. 2856. But an agency confronted with this situation cannot "merely recite the terms 'substantial uncertainty' as a justification for its actions." Ibid. Instead, it "must explain the evidence which is available" and typically must offer a reasoned explanation for taking action without "engaging in a search for further evidence." Ibid.

The Secretary did not do so here. He did not explain why he made the decision to add the question without following the Bureau's ordinary practice of extensively testing proposed changes to the census questionnaire. See App. 624-630, 641 (discussing testing process); see also, e.g. , Brief for Former Census Bureau Directors as Amici Curiae 17-21 (discussing prior examples of questions that the Bureau decided not to add after many years of pretesting). Without that testing, the Secretary could not treat the Bureau's expert opinions and its experience with the relevant surveys as worthless merely because its conclusions were not precise. The Bureau's opinions were properly considered as evidence of likelihoods, probabilities, or risks.

As noted above, the consequences of mistakes in the census count, of even a few hundred thousand, are grave. Differences of a few thousand people, as between one State and another, can mean a loss or gain of a congressional seat-a matter of great consequence to a State. See 351 F.Supp.3d at 594. And similar small differences can make a large difference to the allocation of federal funds among competing state programs. Id. , at 596-597 ; see also Baldrige , 455 U.S. at 353-354, n. 9, 102 S.Ct. 1103. If near-absolute certainty is what the Secretary meant by "definitive," that insistence would itself be arbitrary in light of the constitutional and statutory consequences at stake. And if the Secretary instead meant that the evidence does not indicate a serious risk of a less accurate count, that conclusion does not find support in the record.

B

Now consider the Secretary's conclusion that, even if adding a citizenship question diminishes the accuracy of the enumeration, "the value of more complete and accurate data derived from surveying the entire population outweighs ... concerns" about diminished accuracy. App. to Pet. for Cert. 562a (emphasis added). That conclusion was also arbitrary. The administrative record indicates that adding a citizenship question to the short form would produce less "complete and accurate data," not more.

1

The Census Bureau informed the Secretary that, for about 90% of the population, accurate citizenship data is available from administrative records maintained by the Social Security Administration and Internal Revenue Service. App. 146. The Bureau *2591further informed the Secretary that it had "high confidence" that it could develop a statistical model that would accurately impute citizenship status for the remaining 10% of the population. Ibid. The Bureau stated that these methods alone-using existing administrative records for 90% of the population and statistical modeling for the remaining 10%-would yield more accurate citizenship data than also asking a citizenship question. Id. , at 159. How could that be so? The answer is somewhat technical but readily understandable.

First , consider the 90% of the population (about 295 million people) as to whom administrative records are available. The Government agrees that using these administrative records would provide highly reliable information about citizenship, because the records "require proof of citizenship." Id. , at 117. By contrast, if responses to a citizenship question were used for this group, the Census Bureau predicted without contradiction that about one-third of the noncitizens in this group who respond would answer the question untruthfully, claiming to be citizens when they are not. Id. , at 147. Those incorrect answers-about 9.5 million in total-would conflict with the administrative records on file for those noncitizens. And what would the Census Bureau do with the conflicting data? If it accepts the answer to the citizenship question as determinative, it will have less accurate data. If it accepts the citizenship data from administrative records as determinative, asking the question will have served no purpose.

Thus, as to 295 million people-the overwhelming majority of the population-asking the citizenship question would at best add nothing at all. I say "at best" because, for one thing, the Census Bureau informed the Secretary that asking the question would produce 1 million more people who could not be linked to administrative records, which in turn would require the Census Bureau to resort to a less accurate source of citizenship data for these people. See id. , at 147-149; see also 351 F.Supp.3d at 538-539. For another, the policy of the Census Bureau has always been to use census responses rather than administrative records in cases where the two conflict. App. 147. In this case, that practice would mean accepting 9.5 million inaccurate responses even though accurate administrative records are available. See ibid. The Census Bureau could perhaps change that practice, but the Secretary's decision memorandum said nothing about the matter. It did not address the problem.

Second , consider the remaining 10% of the population (about 35 million people) for whom the Government lacks administrative records. The question here is which approach would yield the most "complete and accurate" citizenship data for this group-adding a citizenship question or using statistical modeling alone? To answer this question, we must further divide this group into two categories-those who would respond to the citizenship question if it were asked and those who would not.

Start with the category of about 22 million people who would answer a citizenship question if it were asked. Would their answers regarding citizenship be more accurate than citizenship data produced by statistical modeling? The Census Bureau said no. That is because many of the noncitizens in this group would answer the question falsely, resulting in an estimated 500,000 inaccurate answers. See id. , at 148. And those who answer the question falsely would be commingled, perhaps randomly, with those who answer it correctly, thereby casting doubt on the answers of all 22 million, with no way of knowing which answers are correct and which are false. By contrast, the Bureau believed that it could develop a statistical model that *2592would produce more accurate citizenship data than these census responses. The Bureau therefore informed the Secretary that it could do better. As the Bureau's chief scientist explained, although "[o]ne might think" that asking the question "could help fill the ... gaps" in the administrative records, the data did not support that assumption. Id. , at 157. Instead, he explained, responses to the citizenship question "may not be reliable," which "calls into question their ability to improve upon" the Bureau's statistical modeling process. Ibid.

Next, turn to the more than 13 million remaining people who would not answer the citizenship question even if it were asked. As to this category, the Census Bureau would still need to use statistical modeling to obtain citizenship data, because there would be no census response to use instead. Hence, asking the citizenship question would add nothing at all as to this group. To the contrary, as the Government concedes, asking the question would reduce the accuracy of the citizenship data for this group, because the relatively inaccurate answers to the citizenship question would diminish the overall accuracy of the Census Bureau's statistical model. See Brief for Petitioners 34 (conceding that the Census Bureau model will be "highe[r] quality" without the question than with it); 351 F.Supp.3d at 640 (explaining that asking the question would "corrup[t] ... the data generated by extrapolating from self-responses through imputation").

In sum, in respect to the 295 million persons for whom administrative records exist, asking the question on the short form would, at best, be no improvement over using administrative records alone. And in respect to the remaining 35 million people for whom no administrative records exist, asking the question would be no better, and in some respects would be worse, than using statistical modeling. The Census Bureau therefore told the Secretary that asking the citizenship question, even in addition to using administrative records, "would result in poorer quality citizenship data" than using administrative records alone, and would "still have all the negative cost and quality implications" of asking the citizenship question. App. 159. I could find no evidence contradicting that prediction.

2

If my description of the record is correct, it raises a serious legal problem. How can an agency support the decision to add a question to the short form, thereby risking a significant undercount of the population, on the ground that it will improve the accuracy of citizenship data, when in fact the evidence indicates that adding the question will harm the accuracy of citizenship data? Of course it cannot. But, as I have just said, I have not been able to find evidence to suggest that adding the question would result in more accurate citizenship data. Neither could the District Court. After reviewing the record in detail, the District Court found that "all of the relevant evidence before Secretary Ross-all of it-demonstrated that using administrative records ... would actually produce more accurate [citizenship] data than adding a citizenship question to the census." 351 F.Supp.3d at 650.

What consideration did the Secretary give to this problem? He stated simply that "[a]sking the citizenship question of 100 percent of the population gives each respondent the opportunity to provide an answer," which "may eliminate the need for the Census Bureau to have to impute an answer for millions of people." App. to Pet. for Cert. 556a. He therefore must have assumed, sub silentio , exactly what *2593the Census Bureau experts urged him not to assume-that answers to the citizenship question would be more accurate than statistical modeling. And he ignored the undisputed respects in which asking the question would make the existing data less accurate. Other than his assumption, the Secretary said nothing, absolutely nothing, to suggest a reasoned basis for disagreeing with the Bureau's expert statistical judgment.

The Government now maintains that the Secretary reasonably discounted the Census Bureau's recommendation because it was based on an untested prediction about the accuracy of its model. But this is not a case in which the Secretary was presented with a policy choice between two reasonable but uncertain options. For one thing, the record is much less uncertain than the Government acknowledges. Although it is true that the Census Bureau at one point told the Secretary that it could not "quantify the relative magnitude of the errors across the alternatives at this time," App. 148, it unequivocally stated that asking the question "would result in poorer quality citizenship data " than omitting it, id. , at 159 (emphasis added). Thus, even if the Bureau could not "quantify" the relative accuracy of the options, it could and did conclude that one option was likely more accurate than the other. Even in the face of some uncertainty, where all available evidence indicates that one option is better than the other, it is unreasonable to choose the worse option without explanation.

For another thing, to the extent the record reflects some uncertainty regarding the accuracy of the Census Bureau's statistical model, that is because the model needed to be "developed and tested" before it could be employed. Id. , at 146. But the Secretary made his decision before any such development or testing could be completed. Having decided to make an immediate decision rather than wait for testing, the Secretary could not dismiss the Bureau's prediction about the inadvisability of that decision on the ground that the prediction reflected likelihoods, probabilities, and risks rather than certainties.

Finally, recall that the Census Act requires the Secretary to use administrative records rather than direct inquiries to "the maximum extent possible." 13 U. S. C. § 6(c). That statutory requirement highlights what should be obvious: Whether adding a citizenship question to the short form would produce more accurate citizenship data is a relevant factor-indeed, a critically important factor-that the Secretary was required to consider. Here, the Secretary did not adequately explain why he rejected the evidence that adding the question would yield less accurate data. He did not even acknowledge that the Census Act obliged him to use administrative records rather than asking a question to the extent possible. And he did not explain how obtaining citizenship data that is no better or worse than the data otherwise available could justify jeopardizing the accuracy of the census count.

In these respects, the Secretary failed to consider "important aspect[s] of the problem" and "offered an explanation for [his] decision that runs counter to the evidence before the agency." State Farm , 463 U.S. at 43, 103 S.Ct. 2856.

C

The Secretary's failure to consider this evidence-that adding the question would harm the census count in the interest of obtaining less accurate citizenship data-provides a sufficient basis for setting the decision aside. But there is more. The reason that the Secretary provided for needing more accurate citizenship information in the first place-to help the DOJ

*2594enforce the Voting Rights Act-is unconvincing.

The Secretary stated that adding the citizenship question was "necessary to provide complete and accurate data in response to the DOJ request." App. to Pet. for Cert. 562a. The DOJ's request in turn asserted that the citizenship data currently available from the ACS was not "ideal" for enforcing the Voting Rights Act. Id. , at 567a. One of the DOJ's principal complaints was that ACS data is reported for groups of census blocks rather than for each census block itself. The DOJ letter stated that adding a citizenship question could provide it with individual block-by-block data which, the DOJ maintained, would allow it to better enforce the Voting Rights Act's protections for minority voters. Id. , at 568a.

This rationale is difficult to accept. One obvious problem is that the DOJ provided no basis to believe that more precise data would in fact help with Voting Rights Act enforcement. Congress enacted the Voting Rights Act in 1965-15 years after the census last asked every household about citizenship. Actions to enforce the Act have therefore always used citizenship data derived from sampling. Yet I am aware of no one-not in the Department of Commerce proceeding, in the District Court, or in this Court-who has provided a single example in which enforcement of the Act has suffered due to lack of more precise citizenship data. Organizations with expertise in this area tell us that asking the citizenship question will not help enforce the Act. See, e.g. , Brief for NAACP Legal Defense & Educational Fund, Inc., as Amicus Curiae 30-36. Rather, the question will, by depressing the count of minority groups, hurt those whom the Act seeks to help. See, e.g. , Brief for Leadership Conference on Civil and Human Rights et al. as Amici Curiae 21-29.

Another problem with the Secretary's rationale is that, even assuming the DOJ needed more detailed citizenship data, there were better ways of obtaining the needed data. The Census Bureau offered to provide the DOJ with data using administrative records, which, as I have pointed out, are likely just as accurate, if not more accurate, than responses to a citizenship question. The Census Bureau offered to provide this data at the census block level, which would resolve each of the DOJ's complaints about the existing ACS data. See Administrative Record 3289. But the Secretary rejected this alternative without explaining why it would not fully respond to the DOJ's request. That failure was particularly problematic given that the Census Act requires the Secretary to use other methods of obtaining demographic information if at all possible. See §§ 6(c), 195.

Normally, the Secretary would be entitled to place considerable weight upon the DOJ's expertise in matters involving the Voting Rights Act, but there are strong reasons for discounting that expertise here. The administrative record shows that DOJ's request to add a citizenship question originated not with the DOJ, but with the Secretary himself. See Administrative Record 3710. The Voting Rights Act rationale was in fact first proposed by Commerce Department officials. See ibid. DOJ officials, for their part, were initially uninterested in obtaining more detailed citizenship data, App. 414, and they agreed to request the data only after the Secretary personally spoke to the Attorney General about the matter, see Administrative Record 2651. And when the acting director of the Census Bureau proposed alternative means of obtaining better citizenship data, DOJ officials declined to meet to discuss the proposal. See id. , at 3460.

*2595Taken as a whole, the evidence in the administrative record indicates that the Voting Rights Act rationale offered by the Secretary was not just unconvincing, but pretextual. And, as the Court concludes, further evidence outside the administrative record but present in the trial record supports the finding of pretext. See Part V, ante. Among other things, that evidence reveals that the DOJ official who wrote the letter agreed that adding the question "is not necessary for DOJ's VRA enforcement efforts." App. 1113. And that official further acknowledged that he did not "know whether or not [citizenship] data produced from responses to the citizenship question ... will, in fact, be more precise than the [citizenship] data on which the DOJ is currently relying for purposes of VRA enforcement." Id. , at 1102.

The Court explains, and I agree, that a court normally should not "reject an agency's stated reasons for acting simply because the agency might also have had other unstated reasons." Ante , at 2573. But in this case, "the evidence tells a story that does not match the explanation the Secretary gave for his decision." Ante , at 2575. This evidence strongly suggests that the Secretary's stated rationale was pretextual. I consequently join Part V of the Court's opinion (except insofar as it concludes that the Secretary's decision was reasonable apart from the question of pretext). And I agree that the pretextual nature of the Secretary's decision provides a sufficient basis to affirm the District Court's decision to send the matter back to the agency.

* * *

I agree with the Court that the APA gives agencies broad leeway to carry out their legislatively delegated duties. And I recognize that Congress has specifically delegated to the Secretary of Commerce the authority to conduct a census of the population "in such form and content as he may determine." § 141(a). But although this delegation is broad, it is not without limits. The APA supplies one such limit. In an effort to ensure rational decisionmaking, the APA prohibits an agency from making decisions that are "arbitrary, capricious, [or] an abuse of discretion." 5 U. S. C. § 706(2)(A).

This provision, of course, does not insist that decisionmakers think through every minor aspect of every problem that they face. But here, the Secretary's decision was a major one, potentially affecting the proper workings of our democratic government and the proper allocation of hundreds of billions of dollars in federal funds. Cf. ante , at 2565 - 2566. Yet the decision was ill considered in a number of critically important respects. The Secretary did not give adequate consideration to issues that should have been central to his judgment, such as the high likelihood of an undercount, the low likelihood that a question would yield more accurate citizenship data, and the apparent lack of any need for more accurate citizenship data to begin with. The Secretary's failures in considering those critical issues make his decision unreasonable. They are the kinds of failures for which, in my view, the APA's arbitrary and capricious provision was written.

As I have said, I agree with the Court's conclusion as to pretext and with the decision to send the matter back to the agency. I do not agree, however, with several of the Court's conclusions concerning application of the arbitrary and capricious standard. In my view, the Secretary's decision-whether pretextual or not-was arbitrary, capricious, and an abuse of his lawfully delegated discretion. I consequently concur in the Court's judgment to the extent that it affirms the judgment of the District Court.

Justice ALITO, concurring in part and dissenting in part.

*2596It is a sign of our time that the inclusion of a question about citizenship on the census has become a subject of bitter public controversy and has led to today's regrettable decision. While the decision to place such a question on the 2020 census questionnaire is attacked as racist, there is a broad international consensus that inquiring about citizenship on a census is not just appropriate but advisable. No one disputes that it is important to know how many inhabitants of this country are citizens.1 And the most direct way to gather this information is to ask for it in a census. The United Nations recommends that a census inquire about citizenship,2 and many countries do so.3

Asking about citizenship on the census also has a rich history in our country. Every census, from the very first one in 1790 to the most recent in 2010, has sought not just a count of the number of inhabitants but also varying amounts of additional demographic information. In 1800, Thomas Jefferson, as president of the American Philosophical Society, signed a letter to Congress asking for the inclusion on the census of questions regarding " 'the respective numbers of native citizens, citizens of foreign birth, and of aliens' " " 'for the purpose ... of more exactly distinguishing the increase of population by birth and immigration.' " C. Wright, History and Growth of the United States Census (prepared for the Senate Committee on the Census), S. Doc. No. 194, 56th Cong., 1st Sess., 19 (1900). In 1820, John Quincy Adams, as Secretary of State, was responsible for conducting the census, and consistent with the 1820 Census Act, he instructed the marshals who were charged with gathering the information to ask about citizenship.4 In 1830, when Martin Van Buren was Secretary of State, a question about citizenship was again included.5 With the exception of the census of 1840, at least some portion of the population was asked a question about citizenship as part of the census through 2000, after which the question was moved to the American Community Survey, which is sent to only a small fraction of the population. All these census inquiries were made by the Executive pursuant to congressional authorization. None were reviewed by the courts.

Now, for the first time, this Court has seen fit to claim a role with respect to the inclusion of a citizenship question on the census, and in doing so, the Court has set a dangerous precedent, both with regard *2597to the census itself and with regard to judicial review of all other executive agency actions. For the reasons ably stated by Justice THOMAS, see ante , p. ---- (opinion concurring in part and dissenting in part), today's decision is either an aberration or a license for widespread judicial inquiry into the motivations of Executive Branch officials. If this case is taken as a model, then any one of the approximately 1,000 district court judges in this country, upon receiving information that a controversial agency decision might have been motivated by some unstated consideration, may order the questioning of Cabinet officers and other high-ranking Executive Branch officials, and the judge may then pass judgment on whether the decision was pretextual. What Bismarck is reputed to have said about laws and sausages comes to mind. And that goes for decisionmaking by all three branches.

To put the point bluntly, the Federal Judiciary has no authority to stick its nose into the question whether it is good policy to include a citizenship question on the census or whether the reasons given by Secretary Ross for that decision were his only reasons or his real reasons. Of course, we may determine whether the decision is constitutional. But under the considerations that typically guide this Court in the exercise of its power of judicial review of agency action, we have no authority to decide whether the Secretary's decision was rendered in compliance with the Administrative Procedure Act (APA).

I

The APA authorizes judicial review of "agency action" taken in violation of law, 5 U. S. C. §§ 706(2)(A)-(D), but § 701(a)(2) of the APA bars judicial review of agency actions that are "committed to agency discretion by law." Although we have characterized the scope of § 701(a)(2) as " 'narrow,' " Heckler v. Chaney , 470 U.S. 821, 830, 105 S.Ct. 1649, 84 L.Ed.2d 714 (1985), there are circumstances in which it applies. And while our cases recognize a strong presumption in favor of judicial review of agency action, see, e.g. , Weyerhaeuser Co. v. United States Fish and Wildlife Serv. , 586 U. S. ----, ----, 139 S.Ct. 361, 370, 202 L.Ed.2d 269 (2018), this "is 'just' a presumption," and like all real presumptions, it may be (and has been) rebutted, Lincoln v. Vigil , 508 U.S. 182, 190, 113 S.Ct. 2024, 124 L.Ed.2d 101 (1993).6

In considering whether the general presumption in favor of judicial review has been rebutted in specific cases, we have identified factors that are relevant to the inquiry: whether the text and structure of the relevant statutes leave a court with any " 'meaningful standard against which to judge the agency's exercise of discretion,' " Webster v. Doe , 486 U.S. 592, 600, 108 S.Ct. 2047, 100 L.Ed.2d 632 (1988) (quoting Heckler , supra , at 830, 105 S.Ct. 1649 ); whether the matter at hand has traditionally been viewed as committed to agency discretion, see ICC v. Locomotive Engineers , 482 U.S. 270, 282, 107 S.Ct. 2360, 96 L.Ed.2d 222 (1987) ; whether the challenged action manifests a "general unsuitability" for judicial review because it involves a "complicated balancing of a number of factors," including judgments regarding the allocation of agency resources or matters otherwise committed to *2598another branch, Heckler , supra , at 831-832, 105 S.Ct. 1649 ; and whether judicial review would produce "disruptive practical consequences," Southern R. Co. v. Seaboard Allied Milling Corp. , 442 U.S. 444, 457, 99 S.Ct. 2388, 60 L.Ed.2d 1017 (1979) (applying this factor to the reviewability inquiry under § 701(a)(1) ).

Applying those factors, I conclude that the decision of the Secretary of Commerce to add core demographic questions to the decennial census questionnaire is committed to agency discretion by law and therefore may not be challenged under the APA.7

II

A

I start with the question whether the relevant statutory provisions provide any standard that courts can apply in reviewing the Secretary's decision to restore a citizenship question to the census. The provision that directly addresses this question is 13 U. S. C. § 141(a), the statute that vests the Secretary with authority to administer the decennial census. This provision gives the Secretary unfettered discretion to include on the census questions about basic demographic characteristics like citizenship. It begins by providing that the Secretary

"shall, in the year 1980 and every 10 years thereafter, take a decennial census of population ... in such form and content as he may determine , including the use of sampling procedures and special surveys." Ibid. (emphasis added).

The two phrases I have highlighted-"census of population" and "in such form and content as he may determine"-are of immediate importance. A "census of population" is broader than a mere head count. The term is defined as "a census of population ... and matters relating to population ." § 141(g) (emphasis added). Because this definition refers to both "a census of population" and "matters relating to population," the latter concept must include more than a "census of population" in the strict sense of a head count. And it seems obvious that what this additional information must include is the sort of basic demographic information that has long been sought in the census. So the statute clearly authorizes the Secretary to gather such information.

The second phrase, "in such form and content as he may determine," specifies how this information is to be gathered, namely, by a method having the "form and content" that the Secretary "may determine." In other words, this is left purely to the Secretary's discretion. A clearer and less restricted conferral of discretion is hard to imagine.

It is instructive to compare this delegation of authority to the statutory language at issue in one of our most well-known § 701(a)(2) cases, Webster v. Doe , 486 U.S. 592, 108 S.Ct. 2047, 100 L.Ed.2d 632. There, the relevant statute allowed termination of a Central Intelligence Agency employee whenever the Director "shall deem such termination necessary or advisable in the interests of the United States." Id. , at 600, 108 S.Ct. 2047 (internal quotation marks omitted and emphasis deleted). Reasoning that the statute's "shall deem " standard "fairly exudes *2599deference to the Director," the Court concluded that the text of the statute "appear[ed] ... to foreclose the application of any meaningful judicial standard of review." Ibid.

The § 141(a) language discussed above is even more sweeping than that of the statute in Webster. Unlike the Census Act, the statute in Webster placed a condition on the Director's action-in particular, the requirement that he terminate an employee only after concluding that doing so would further the "interests of the United States." No such condition applies to the Secretary's determination about the form and content of the decennial census, a fact that distinguishes the statute at issue here from others this Court has found to fall outside § 701(a)(2) and thus within courts' power to review. See, e.g., Weyerhaeuser Co. , 586 U. S., at ----, 139 S.Ct., at 370 (statute conditioning agency power to exclude land from critical habitat designation on agency's consideration of " 'economic impact' " of designation and " 'determin[ation] that the benefits of such exclusion outweigh the benefits of specifying such area as part of the critical habitat' ").

B

Those arguing in favor of judicial review contend that the § 141(a) language that I have discussed so far is limited by language that follows immediately after. That part of § 141(a) states:

"In connection with any such census [i.e. , the decennial "census of population"], the Secretary is authorized to obtain such other census information as necessary ." (Emphasis added.)

This means, it is argued, that information about citizenship may be obtained by means of the census only if that is "necessary." But this argument is clearly wrong. The information that must be "necessary" (whatever that means in this context) is "other census information." That refers to information other than that obtained in the "census of population," and as explained, the term "census of population" includes not just a head count but other "matters relating to population," a category that encompasses basic demographic information such as citizenship. Accordingly, this argument is definitively refuted by the text of § 141. And although it is not necessary to look beyond that text, it is worth noting that this argument, if accepted, would require that the term "necessary" be given a less than strictly literal meaning; otherwise, it would run contrary to the broad delegation effected by the first portion of § 141(a) by making it all but impossible for the Secretary to include on the census anything other than questions relating to the number of persons living at a particular address. That would be so because it will often not be "necessary" to obtain this information via the census rather than by some other means.

C

Another argument in favor of review relies on 13 U. S. C. § 195, which states:

"Except for the determination of population for purposes of apportionment of Representatives in Congress among the several States, the Secretary shall, if he considers it feasible, authorize the use of the statistical method known as 'sampling' in carrying out the provisions of this title."

Justice BREYER, for example, interprets this provision to mean that "the Secretary must, if feasible, obtain demographic information through a survey sent to a sample of households, rather than through the short-form census questionnaire to which every household must respond." Ante , at 2585 (opinion concurring in part and dissenting in part). Under that reading of § 195, it is asserted, the provision sets *2600forth a judicially reviewable limit on the Secretary's authority to obtain information through direct inquiries.

This argument fails to take into account that the current version of § 195 was enacted as part of the same Act of Congress that included the present version of § 1418 and that the two provisions are both parts of a unified scheme regarding the use of sampling. Section 141, a provision concerned exclusively with the census, addresses the use of sampling in that particular context. I previously quoted the relevant language, but I repeat it now so that it is clearly in mind. Section 141(a) provides that the Secretary

"shall, in the year 1980 and every 10 years thereafter, take a decennial census of population ... in such form and content as he may determine, including the use of sampling procedures and special surveys ." (Emphasis added.)

What this means is that the Secretary, in conducting the "census of population," has discretion to choose the form and content of the vehicles used in that project, and among the methods that he may employ, if he sees fit, are sampling and special surveys.

Section 195 is not a census-specific provision, but it does have one (important) thing to say specifically about the census: It prohibits the use of sampling "for the determination of population for purposes of apportionment of Representatives in Congress." In this one way, it qualifies the Secretary's discretion regarding the "form and content" of the vehicles used in conducting the "census of population." And that is what we meant in Department of Commerce v. United States House of Representatives , 525 U.S. 316, 338, 119 S.Ct. 765, 142 L.Ed.2d 797 (1999), when we said that § 141(a) 's "broad grant of authority ... is informed ... by the narrower and more specific § 195." Otherwise, the text of § 195 does not deal specifically with the census. It addresses all the many information-gathering activities conducted by the Commerce Department, and as to these, it says that the Secretary shall use sampling if he deems it "feasible."

If § 195 were read to mean that no information other than a head count can be sought by means of a census questionnaire unless it is not "feasible" to get that information by sampling, then there would be little if anything left of the broad discretion "to use sampling techniques" conferred on the Secretary by § 141(a). "Feasible" means "capable of being done, executed, or effected," Webster's Third New International Dictionary 831 (1961), and it is not clear that the gathering of any core demographic information is not "capable of being done" by sampling. So if that were what § 195 means, then Congress, in the same Act, would have given the Secretary discretion to use sampling in the census "as he may determine" but also compelled him to use sampling in almost all instances. That is no way to read the provisions of a single Act. A law's provisions should be read to work together. See A. Scalia & B. Garner, Reading Law 180 (2012) ("The provisions of a text should be interpreted in a way that renders them compatible, not contradictory"). See also, e.g. , Parker Drilling Management Services, Ltd. v. Newton , 587 U. S. ----, ---- - ----, 139 S.Ct. 1881, 1887-1889, L.Ed.2d ---- (2019) (slip op., at 5-6) ; Star Athletica, L. L. C. v. Varsity Brands, Inc. , 580 U. S. ----, ---- - ----, 137 S.Ct. 1002, 1009-1010, 197 L.Ed.2d 354 (2017) ; Kawasaki Kisen Kaisha Ltd. v. Regal-Beloit Corp. , 561 U.S. 89, 108, 130 S.Ct. 2433, 177 L.Ed.2d 424 (2010). And if there is tension between a specific provision, like *2601§ 141 's instruction regarding the use of sampling in the decennial census, and a general one, like § 195 's directive regarding the use of sampling in all data-collection activities, the specific provision must take precedence. Cf. NLRB v. SW General, Inc. , 580 U. S. ----, ----, 137 S.Ct. 929, 941-942, 197 L.Ed.2d 263 (2017).

When §§ 141 and 195 are read in this way, it is easy to see how they fit together. In using the census to gather information "relating to population" for any use other than the actual enumeration, the Secretary may use sampling "as he may determine." In conducting all the Department's efforts to collect data by other means, he may authorize the use of sampling if he thinks that is "feasible." The upshot for present purposes is that § 195 does not require the "counterintuitive resul[t]" of barring the Secretary from including on the census questionnaire the kinds of basic demographic questions that have been asked as part of every census in U. S. history. RJR Nabisco, Inc. v. European Community , 579 U. S. ----, ----, 136 S.Ct. 2090, 2104, 195 L.Ed.2d 476 (2016).

D

One additional provision, 13 U. S. C. § 6(c),9 requires close consideration. This provision, which was enacted in 1976 in the same Act as §§ 141(a) and 195, has three subsections. Subsection (a) provides that the Secretary may call on other components of the Federal Government to obtain information that is "pertinent to" the Department's work. Subsection (b) authorizes the Secretary to "acquire, by purchase or otherwise" from state and local governments and private sources "such copies of records, reports, and other material as may be required for the efficient and economical conduct of the censuses and surveys provided for in this title." Finally, subsection (c) provides:

"To the maximum extent possible and consistent with the kind, timeliness, quality and scope of the statistics required, the Secretary shall acquire and use information available from any source referred to in subsection (a) or (b) of this section instead of conducting direct inquiries."

The District Court interpreted subsection (c) to mean that the Secretary must turn to another federal agency or outside source for demographic information (rather than seeking the information on the census) unless doing so would not be "possible" or "consistent with the kind, timeliness, quality and scope of the statistics required." This argument fails for reasons similar to those that sank the § 195 argument just discussed. Section 6(c) is not a census-specific provision but instead applies generally to all the Commerce Department's information-gathering activities. If it is read to apply to the "census of population," it cannot be reconciled with § 141(a), which, as noted, broadly authorizes the Secretary to use that vehicle for obtaining information "relating to population,"

*2602i.e. , core demographic information. If § 6(c) applied to the gathering of such information, it would make it hard to justify the inclusion of any demographic questions on the census, even though this has been done since 1790. (Is it not possible to get information about age and sex, for example, from any outside source (or combination of sources), even if the Department offers to acquire it from a private source by purchase?) Reading § 6(c) to mean what the District Court thought would turn it into the proverbial elephant stuffed into a mouse hole. Section 6(c), however, is a decidedly mouse-like provision. It was enacted with no fanfare and no real explanation,10 and remained in the shadows, virtually unused and unnoticed, for more than 40 years.

E

Respondents and the Court cite two other provisions in support of reviewability, but neither has anything to do with the issue of putting a citizenship question on the census. In determining whether statutory provisions include standards that could provide a basis for judicial review, it is necessary to focus on the precise claims at issue, see, e.g., Webster , 486 U.S. at 601-602, 108 S.Ct. 2047 (distinguishing between statutory and constitutional claims); Locomotive Engineers , 482 U.S. at 277-279, 107 S.Ct. 2360 (parsing claims under different prongs of reopener statute); Heckler , 470 U.S. at 836, 105 S.Ct. 1649 (rejecting as "irrelevant" to the agency decision at issue two statutory provisions that were argued to provide " 'law to apply' "). And when viewed in this way, the remaining statutory provisions cited in support of reviewability are of no value.

Respondents point to § 141(b), which requires the Secretary to complete the tabulation of total population by States "within 9 months after the census date" and then to report the results to the President. That provision sets out an easily administered deadline, and it has nothing to do with the content of the census questionnaire.

Respondents also claim that § 141(f) is relevant to the question of judicial review, but that provision concerns congressional review. It directs the Secretary to report to Congress, at specified times, the subjects and questions that he intends to include on the census. According to respondents, the Secretary's compliance with those requirements is judicially reviewable, and that, they contend, takes the Secretary's decision to include a citizenship question out from under § 701(a)(2).

Respondents fundamentally misunderstand the significance of congressional reporting requirements in evaluating whether a particular agency action is subject to judicial review. Congressional reporting requirements are "legion in federal law," Natural Resources Defense Council, Inc. v. Hodel , 865 F.2d 288, 317 (CADC 1988), and their purpose is to permit Congress to monitor and, if it sees fit, to correct Executive Branch actions to which it objects. When a congressional reporting requirement "[l]ack[s] a provision for judicial review," compliance "by its nature seems singularly committed to congressional discretion in measuring the fidelity of the *2603Executive Branch actor to legislatively mandated requirements." Id. , at 318. In other words, it is Congress, not the Judiciary, that is best situated to determine whether an agency's responses to Congress are sufficient and, if not, to "take what it deems to be the appropriate action." Id. , at 319.

In that respect, § 141(f) actually cuts against judicial review. The Constitution gives Congress the authority to "direct" the "Manner" in which the census is conducted, and by imposing the § 141(f) reporting requirements, Congress retained some of that supervisory authority. It did not transfer it to the courts.11

Respondents protest that congressional review may not be enough to guard against a Secretary's abuses, especially when the party in control of Congress stands to benefit. But that complaint simply expresses disagreement with the Framers' choice to vest power over the census in a political body, cf. Baldrige v. Shapiro , 455 U.S. 345, 347-348, 102 S.Ct. 1103, 71 L.Ed.2d 199 (1982) ("Under [the] Constitution, responsibility for conducting the decennial census rests with Congress"), and the manner in which Congress has chosen to exercise that power, see Wisconsin v. City of New York , 517 U.S. 1, 19, 116 S.Ct. 1091, 134 L.Ed.2d 167 (1996) (Congress has delegated its "virtually unlimited discretion" in conducting the census to the Secretary). In any event, the ability to press constitutional challenges to the Secretary's decisions, see n. 7, supra, answers many of the examples in respondents' parade of horribles.

In short, the relevant text of § 141(a) "fairly exudes deference" to the Secretary. Webster , 486 U.S. at 600, 108 S.Ct. 2047. And no other provision of law cited by respondents or my colleagues provides any "meaningful judicial standard" for reviewing the Secretary's selection of demographic questions for inclusion on the census. Ibid.

III

In addition to requiring an examination of the text and structure of the relevant statutes, our APA § 701(a)(2) cases look to whether the agency action in question is a type that has traditionally been viewed as committed to agency discretion or whether it is instead one that "federal courts regularly review." Weyerhaeuser Co. , 586 U. S., at ----, 139 S.Ct., at 370. In cases where the Court has found that agency action is committed to agency discretion by law, an important factor has been the absence of an established record of judicial review prior to the adoption of the APA. See Heckler , 470 U.S. at 832-833, 105 S.Ct. 1649 (agency nonenforcement); Locomotive Engineers , 482 U.S. at 282, 107 S.Ct. 2360 (agency decision not to reopen final decision based on material error); Lincoln , 508 U.S. at 192, 113 S.Ct. 2024 (agency use of lump-sum appropriations).

*2604Here, there is no relevant record of judicial review. We are confronted with a practice that reaches back two centuries. The very first census went beyond a mere head count and gathered additional demographic information, and during virtually the entire period prior to the enactment of the APA, a citizenship question was asked of everyone. Notably absent from that long record is any practice of judicial review of the content of the census. Indeed, this Court has never before encountered a direct challenge to a census question. App. to Pet. for Cert. 416a. And litigation in the lower courts about the census is sparse and generally of relatively recent vintage.

Not only is this sort of history significant in all § 701(a)(2) cases, see Locomotive Engineers , supra , at 282, 107 S.Ct. 2360, but we have previously stressed the particular "importance of historical practice" when it comes to evaluating the Secretary's authority over the census. Wisconsin , supra , at 21, 116 S.Ct. 1091 ; see also ante, at 2567 (opinion of the Court). Moreover, where the relevant question is not whether review may be had at all, but rather the branch with the authority to exercise review, the absence of any substantial record of judicial review is especially revealing. See, e.g., NLRB v. Noel Canning , 573 U.S. 513, 525, 134 S.Ct. 2550, 189 L.Ed.2d 538 (2014) (it is "neither new nor controversial" that "longstanding practice of the government can inform our determination of what the law is" (internal quotation marks and citation omitted)); United States v. Midwest Oil Co. , 236 U.S. 459, 473, 35 S.Ct. 309, 59 L.Ed. 673 (1915) ("in determining ... the existence of a power, weight [is] given to ... usage"). Thus, the absence of any real tradition of judicial review of decisions regarding the content of the census counsels against review in this case.

In an attempt to show that there is no relevant "tradition of nonreviewability," Locomotive Engineers , supra , at 282, 107 S.Ct. 2360, respondents contend that this Court has recently engaged in review of the "conduct of the census," Brief for Government Respondents 26-27. But in none of the cases they cite did the Court address an APA challenge to the content of census questions.12 Some involved constitutional claims about enumeration and apportionment. See Franklin v. Massachusetts , 505 U.S. 788, 790, 801, 112 S.Ct. 2767, 120 L.Ed.2d 636 (1992) (constitutional challenge to "method used for counting federal employees serving overseas" as part of "reapportionment determination"); Wisconsin , 517 U.S. at 20, 116 S.Ct. 1091 (constitutional challenge to Secretary's decision not to adjust count). Others concerned enforcement of statutes with specific directives. See Department of Commerce , 525 U.S. at 343, 119 S.Ct. 765 (holding that § 195 bars use of "sampling" to reach actual enumeration for apportionment); Utah v. Evans , 536 U.S. 452, 464-465, 122 S.Ct. 2191, 153 L.Ed.2d 453 (2002) (considering whether statistical method violated § 195 's bar on use of "sampling" in apportionment enumeration). According to respondents, these cases mean that all the Secretary's census-related decisions are suitable for judicial review and thus fall outside of § 701(a)(2), and the Court apparently agrees, rejecting the Government's § 701(a)(2) argument in part because "[w]e and other courts have entertained both constitutional and statutory challenges to census-related decisionmaking." Ante , at 2568.

*2605This argument misses the point of § 701(a)(2). The question under that provision is whether the challenged action "is committed to agency discretion by law," not whether a different action by the same agency is reviewable under the APA, much less whether an action taken by the same agency can be challenged under the Constitution. Take the example of Heckler v. Chaney , supra , where the Court considered whether a particular Food and Drug Administration (FDA) decision was reviewable under the APA. Many FDA actions are subject to APA review, see, e.g., Weinberger v. Hynson, Westcott & Dunning, Inc. , 412 U.S. 609, 627, 93 S.Ct. 2469, 37 L.Ed.2d 207 (1973), but that did not prevent the Heckler Court from holding that the particular FDA decision at issue there fell within § 701(a)(2). See also, e.g. , Heckler , supra, at 836-837, 105 S.Ct. 1649.

Respondents and some of their amici contend that the Secretary's decision is at least amenable to judicial review for consistency with the APA's reasoned-explanation requirement. See Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co. , 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983) (describing requirement). Thus, the argument goes, even if no statute sets out a standard that can be used in reviewing the particular agency action in question, a court may review an agency's explanation of the reasons for its action and set it aside if the court finds those reasons to be arbitrary or irrational.

This argument would obliterate § 701(a)(2). Even if a statute expressly gave an agency absolute, unrestricted, unfettered, unlimited, and unqualified discretion with respect to a particular decision, a court could still review the agency's explanation of the reasons for its decision. That is not what § 701(a)(2) means. As we put it previously in answering a similar argument against application of § 701(a)(2), it is "fals[e]" to suggest "that if the agency gives a 'reviewable' reason for otherwise unreviewable action, the action becomes reviewable." Locomotive Engineers , 482 U.S. at 283, 107 S.Ct. 2360. That is because when an action "is committed to agency discretion by law," the Judiciary has no role to play, even when an agency sets forth "an eminently 'reviewable' proposition." Id., at 282-283, 107 S.Ct. 2360.

IV

In sum, neither respondents nor my colleagues have been able to identify any relevant, judicially manageable limits on the Secretary's decision to put a core demographic question back on the census. And without an "adequate standard of review for such agency action," id. , at 282, 107 S.Ct. 2360, courts reviewing decisions about the "form and content" of the census would inevitably be drawn into second-guessing the Secretary's assessment of complicated policy tradeoffs,13 another indicator of "general unsuitability" for judicial review. Heckler , supra , at 831, 105 S.Ct. 1649.

Indeed, if this litigation is any indication, widespread judicial review of the Secretary's conduct of the census will usher in an era of "disruptive practical consequences,"

*2606and this too weighs against review. Seaboard Allied Milling Corp. , 442 U.S. at 457, 99 S.Ct. 2388. Cf. Tucker v. United States Dept. of Commerce , 958 F.2d 1411, 1418 (CA7 1992) (expressing doubt about "both the provenance and the practicability" of allowing judicial review of census-related decisions).

Respondents protest that the importance of the census provides a compelling reason to allow APA review. See also ante, at 2595 (opinion of BREYER, J.). But this argument overlooks the fact that the Secretary is accountable in other ways for census-related decisionmaking.14 If the Secretary violates the Constitution or any applicable statutory provision related to the census, his action is reviewable. The Secretary is also accountable to Congress with respect to the administration of the census since he has that power only because Congress has found it appropriate to entrust it to him. And the Secretary is always answerable to the President, who is, in turn, accountable to the people.

* * *

Throughout our Nation's history, the Executive Branch has decided without judicial supervision or interference whether and, if so, in what form the decennial census should inquire about the citizenship of the inhabitants of this country. Whether to put a citizenship question on the 2020 census questionnaire is a question that is committed by law to the discretion of the Secretary of Commerce and is therefore exempt from APA review. The District Court had the authority to decide respondents' constitutional claims, but the remainder of their complaint should have been dismissed.

I join Parts I, II, III, IV-B, and IV-C15 of the opinion of the Court. I do not join the remainder, and insofar as the Court holds that the Secretary's decision is reviewable under the APA, I respectfully dissent.

Vacated and remanded by published opinion. Judge GREGORY wrote the opinion, in which Judge DAVIS and Judge WYNN joined.

GREGORY, Circuit Judge:

In this appeal we review the application of a state prison’s policy conditioning an inmate’s request for a religious accommodation on his possession of physical indicia of faith. We also address whether the inmate’s claims for equitable relief were mooted after the prison abandoned the policy. We vacate the district court’s summary judgment order granting the defendants qualified immunity on the plaintiffs claims for monetary relief. We also vacate the district court’s decision that the prison’s abandonment of the policy mooted the claims for equitable relief. We remand to the district court for such further proceedings as may be appropriate.

I.

The plaintiff, Gary Wall, is a state prisoner housed at Red Onion State Prison (“ROSP”) in Pound, Virginia. As a member of the Nation of Islam, in 2008 and 2009 Wall was allowed to observe the holy month of Ramadan while in state custody. To accommodate Ramadan observance, prison officials provide participating inmates with special meals served before sunrise and after sunset. While at ROSP, Wall also received “common fare” meals, which satisfied his religious beliefs.

Prior to 2010, Muslim inmates at ROSP simply had to sign up to participate in Ramadan. In 2009, approximately half of the inmate population signed up. ROSP staff later determined that a significant number of the participating inmates were not, in fact, practicing Muslims. As a result, ROSP devised a new eligibility policy for 2010: in addition to signing up, inmates had to provide some physical indi-cia of Islamic faith, such as a Quran, Kufi, prayer rug, or written religious material obtained from the prison Chaplain’s office. 1 ROSP inmates who did not have such materials or refused to acquire them were deemed insincere in their religious beliefs and were prohibited from participating in Ramadan. 2

*495 Wall was one of the inmates who was not allowed to participate. After initially signing up, Wall was asked by defendants James Wade, C. Selyers, and J. Stallard to provide physical evidence of the sincerity of his beliefs in accordance with the new policy. Wall stated that all his belongings, including his articles of faith, had been lost during his transfer to ROSP. He showed Wade a state court judgment against the Commonwealth as proof that VDOC had lost his possessions. 3 Wall also produced documents showing that he was receiving common fare meals in accordance with his faith, and he informed the officers that he had observed Ramadan in 2008 and 2009. Despite this, Wade responded, “that don’t mean anything,” and instructed Stallard and Selyers to remove Wall from the Ramadan list. J.A. 139.

Wall then filed an informal complaint, again explaining that his religious materials had been lost and requesting to be allowed to participate. In a memo in response to the complaint, Wade reiterated ROSP’s new policy, stating:

[ROSP] does not have religious services so the following rules apply to this institution. You are required to have religious material such as ([ku]f[i], [Qu]r[a] n, prayer rug or religious pamphlets that pertain to the Ramadan month long fasting.) Food service went to every inmate[’]s cell to inspect the above religious material. Either you had no religious material or refused to present material!.] [T]his is why you were removed from the Ramadan pass list.

J.A. 42.

On August 11, 2010, the first morning of Ramadan, Wall did not eat breakfast and concealed a portion of his meal in his cell to save until after sunset. ROSP staff found the food and threatened to charge him with possessing contraband. Faced with choosing between starvation and sanctions, Wall ate during the day and violated his religious beliefs.

On August 15, Wall filed a formal grievance, which was also denied. Six days later, he had a conversation with Wade and Assistant Warden Robert Rowlette, in which Rowlette asked if he would like to be put back on the Ramadan list provided it could be verified that he had truly lost his belongings. According to Wall, he responded that he still wanted to participate, but that he also wanted an explanation for why he was taken off the list in the first place. Rowlette replied, “[o]kay,” and then walked away while Wall shouted “I want to participate in Ramadan! I want my Ramadan, Rowlette!” J.A. 140. According to the defendants, however, Wall refused Rowlette’s offer to be put back on the list, saying, “[n]o, I’m going to pursue this in court.” J.A. 93. Ultimately, Wall was not allowed to participate in Ramadan in 2010.

Having exhausted his administrative remedies, Wall filed suit under the Religious Land Use and Institutionalized Persons Act (“RLUIPA”), 42 U.S.C. § 2000cc, et seq., and 42 U.S.C. § 1983. In an amended complaint, Wall alleged violations of RLUIPA and the Free Exercise Clause of the First Amendment, in addition to several related state law claims. The amended complaint sought “[a] declaratory judgment, nominal damages, unspecified joint and several compensatory damages, $10,000 in punitive damages from each defendant, and any additional relief this court deems just, proper, and equitable.” J.A. 32-33.

*496 Shortly after the suit was filed, Wall was transferred out of ROSP to a lower-security facility. Subsequently, the district court granted the defendants’ motion for summary judgment, finding that any claims for equitable relief were moot following Wall’s transfer, and ruling that the defendants were entitled to qualified immunity on the plaintiffs claim for damages.

Following the district court’s ruling, Wall was transferred back to ROSP. The defendants claim that ROSP has since abandoned its policy of requiring prisoners to possess physical indicia of faith in order to participate in Ramadan or other religious observations. The new policy, adopted in a September 13, 2011 memo by VDOC’s Chief of Corrections Operations, 4 states that inmates in segregation facilities, such as ROSP, may demonstrate sincerity by showing that they have in the past borrowed religious material such as DVDs, CDs, or literature from the Chaplain’s office. The memo states that the change was made following an investigation by VDOC’s Inspector General, which concluded that “it is not appropriate to require inmates to buy something which is related to exercising First Amendment [flights.” Following the policy change, Wall and other inmates who were prohibited from observing Ramadan in 2010 were allowed to participate in a “make-up” Ramadan in April 2012.

II.

We review two issues in this appeal: whether the district court correctly determined that Wall’s equitable claims under RLUIPA and the First Amendment were moot following ROSP’s decision to abandon the 2010 Ramadan policy; and whether the district court correctly granted the defendants qualified immunity on Wall’s First Amendment claim for damages. 5 Both issues are questions of law which we review de novo. See Green v. City of Raleigh, 523 F.3d 293, 298 (4th Cir.2008) (mootness); Johnson v. Caudill, 475 F.3d 645, 650 (4th Cir.2007) (qualified immunity). We address the issues in turn.

A.

In granting the defendants’ motion for summary judgment, the district court found that Wall’s transfer to another facility mooted his request for equitable relief. 6 *497 Although Wall’s subsequent return to ROSP rendered this justification obsolete, the district court also ruled that in such an event Wall’s claims would remain moot in light of VDOC’s decision to terminate the 2010 Ramadan policy.

It is well established that a defendant’s “voluntary cessation of a challenged practice” moots an action only if “subsequent events made it absolutely clear that the allegedly wrongful behavior could not reasonably be expected to recur.” Friends of the Earth, Inc. v. Laidlaw Envtl. Servs., Inc., 528 U.S. 167, 189, 120 S.Ct. 693, 145 L.Ed.2d 610 (2000); see Knox v. Service Employees Intern. Union, Local 1000, - U.S. -, 132 S.Ct. 2277, 2287, 183 L.Ed.2d 281 (2012) (“The voluntary cessation of challenged conduct does not ordinarily render a case moot because a dismissal for mootness would permit a resumption of the challenged conduct as soon as the case is dismissed.”). Were it otherwise, “courts would be compelled to leave ‘[t]he defendant ... free to return to his old ways.’ ” City of Mesquite v. Aladdin’s Castle, 455 U.S. 283, 289 n. 10, 102 S.Ct. 1070, 71 L.Ed.2d 152 (1982) (quoting United States v. W.T. Grant Co., 345 U.S. 629, 632, 73 S.Ct. 894, 97 L.Ed. 1303 (1953)). “The ‘heavy burden of per-suafding]’ the court that the challenged conduct cannot reasonably be expected to start up again lies with the party asserting mootness.” Laidlaw, 528 U.S. at 189, 120 S.Ct. 693 (quoting United States v. Concentrated Phosphate Export Ass’n, 393 U.S. 199, 203, 89 S.Ct. 361, 21 L.Ed.2d 344 (1968)).

We have no difficulty concluding that the defendants failed to meet their “heavy burden” of establishing that it is “absolutely clear” the 2010 Ramadan policy will not be reinstated. Id. Unsubstantiated assurances in their appellate brief aside, the defendants have failed to put forth even a single piece of evidence establishing that the practice of requiring physical indicia of faith has been terminated once and for all. The September 13, 2011 memorandum describing VDOC’s purported change in policy&emdash;which was only submitted in a different case 7 &emdash;also fails to establish that VDOC will not reinstate the policy following completion of this lawsuit. We have previously held that when a defendant retains the authority and capacity to repeat an alleged harm, a plaintiffs claims should not be dismissed as moot. Town of Nags Head v. Toloczko, 728 F.3d 391, 395 n. 3 (4th Cir.2013); Pashby v. Delia, 709 F.3d 307, 316 (4th Cir.2013); Lyons P’ship, L.P. v. Morris Costumes, Inc., 243 F.3d 789, 800 (4th Cir.2001). Nothing in the memo suggests that VDOC is actually barred&emdash;or even considers itself barred&emdash;from reinstating the 2010 Ramadan policy should it so choose. To the contrary, the fact that at least three separate policies have been utilized at ROSP since 2009 indicates some degree of doubt that the new policy will remain in place for long.

The defendants invite us to adopt an approach employed by several of our sister circuits, in which governmental defendants are held to a less demanding burden of proof than private defendants. See, e.g., Rio Grande Silvery Minnow v. Bureau of Reclamation, 601 F.3d 1096, 1116 (10th Cir.2010) (“In practice ..., Laidlaw’s heavy burden frequently has not prevented governmental officials from discontinuing challenged practices and mooting a case.”); Sossamon v. Texas, 560 F.3d 316, 325 (5th Cir.2009) (“[Cjourts are justified in treating a voluntary governmental ces *498 sation of possibly wrongful conduct with some solicitude, mooting cases that might have been allowed to proceed had the defendant not been a public entity.”)- However, even if we were to adopt this approach, a question which we expressly do not decide, we would have no trouble determining that the defendants’ near total failure to provide the Court with information regarding the change would remain insufficient even under a lesser standard. In short, bald assertions of a defendant&emdash; whether governmental or private&emdash;that it will not resume a challenged policy fail to satisfy any burden of showing that a claim is moot. 8 We therefore vacate the district court’s dismissal of the plaintiffs equitable claims.

B.

Turning to the plaintiffs claim for monetary damages under the First Amendment, the district court ruled that the defendants were entitled to. qualified immunity. 9 Qualified immunity protects government officials performing discretionary functions unless: “(1) the allegations underlying the claim, if true, substantiate the violation of a federal statutory or constitutional right; and (2) this violation was of a clearly established right of which a reasonable person would have known.” Ridpath v. Bd. of Governors Marshall Univ., 447 F.3d 292, 306 (4th Cir.2006) (internal quotation marks omitted). Viewing the facts in the light most favorable to the plaintiff, we conclude that the defendants have failed to establish as a matter of law that the 2010 Ramadan policy, as applied to the plaintiff, did not violate his First Amendment rights. We also hold that their alleged actions violated the plaintiffs clearly established rights, and that they are therefore not entitled to qualified immunity.

1.

“The Free Exercise Clause of the First Amendment forbids the adoption of laws designed to suppress religious beliefs or practices.” Morrison v. Garraghty, 239 F.3d 648, 656 (4th Cir.2001). This encompasses policies that impose a substantial burden on a prisoner’s right to practice his religion. Lovelace v. Lee, 472 F.3d 174, 198 & n. 8 (4th Cir.2006). “Under ... the Free Exercise Clause ..., a prisoner has a ‘clearly established ... right to a diet consistent with his ... religious scruples,’ including proper food during Ramadan.” Id. at 198-99 (quoting Ford v. McGinnis, 352 F.3d 582, 597 (2nd Cir.2003)). The defendants concede that denying Wall the opportunity to observe Ramadan imposed a substantial burden on his religious freedom.

*499 However, free exercise restrictions that are “reasonably adapted to achieving a legitimate penological objective” are permissible. Id. at 200. In other words, then, prison officials may restrict an inmate’s religious practices subject to a “reasonableness” test that accords substantial deference to the professional judgment of correctional officers. See Overton v. Bazzetta, 539 U.S. 126, 132, 123 S.Ct. 2162, 156 L.Ed.2d 162 (2003). 10

A prison regulation is reasonable and thus permissible if it satisfies the four factors established in Turner v. Safley, 482 U.S. 78, 107 S.Ct. 2254, 96 L.Ed.2d 64 (1987). That test asks: (1) whether there is a “valid, rational connection” between the prison regulation or action and the interest asserted by the government, or whether this interest is “so remote as to render the policy arbitrary or irrational”; (2) whether “alternative means of exercising the right ... remain open to prison inmates”; (3) what impact the desired accommodation would have on security staff, inmates, and the allocation of prison resources; and (4) whether there exist any “obvious, easy alternatives” to the challenged regulation or action. Lovelace, 472 F.3d at 200 (citing Turner, 482 U.S. at 89-92, 107 S.Ct. 2254).

As a preliminary matter, “prison officials may appropriately question whether a prisoner’s religiosity, asserted as the basis for a requested accommodation, is authentic.” Cutter v. Wilkinson, 544 U.S. 709, 725 n. 13, 125 S.Ct. 2113, 161 L.Ed.2d 1020 (2005); see Gillette v. United States, 401 U.S. 437, 457, 91 S.Ct. 828, 28 L.Ed.2d 168 (1971) (“[T]he ‘truth’ of a belief is not open to question; rather, the question is whether the objector’s beliefs are ‘truly held.’”) (internal quotation marks omitted). Accepting that prisons may limit religious accommodations to sincere believers, the question in this case is whether ROSP’s specific means of testing Wall’s sincerity was permissible; that is, whether ROSP was allowed to require him to possess specific, physical items of Islamic faith as proof of belief.

We hold that under the current record, the defendants’ application of the 2010 Ramadan policy fails an analysis under the Turner factors. 11 First, demanding specific physical items as proof of faith will rarely be an acceptable means of achieving the prison’s stated interest in reducing costs. Strict application of such a rule fails even a rational connection requirement. Although we recognize that prison officials must make determinations of who is entitled to accommodations, it exceeds their authority to decide which, if any, religious relics are sufficiently important as to constitute an appropriate gauge of faith. This Court has held, albeit under *500 RLUIPA’s more exacting standard, that prison administrators may not assume that a “lack of sincerity (or religiosity) with respect to one practice [of a given religion] means lack of sincerity with respect to others.” Lovelace, 472 F.3d at 188. Thus, the fact that Wall did not have, for example, a prayer rug in his possession is not a sufficiently reliable indicator of whether he is a practicing Muslim. A prison may not condition an inmate’s constitutionally protected rights on so narrow a set of grounds without “rendering] the policy arbitrary or irrational.” Turner, 482 U.S. at 89-90, 107 S.Ct. 2254.

Indeed, the circumstances of Wall’s case highlight exactly why such an unyielding policy is unreasonable. Despite Wall’s other outward manifestations of faith, most notably his past participation in Ramadan and common fare diet, 12 he was prohibited from observing the fast solely because he did not possess any of the approved items. The defendants also ignored Wall’s perfectly believable explanation, later verified, that he did not have the items only because VDOC had lost all of his belongings during his transfer to ROSP. Nor was Wall’s attempt to store food in his cell on the first day of Ramadan enough to convince ROSP officials that he was sincere in his desire to observe the fasting requirements. Wall also continued to pursue the issue after his initial denial, filing several grievances and requesting to be placed back on the list. The defendants rejected these requests and simply reiterated their policy without further consideration of Wall’s circumstances. Finally, at least according to Wall, on August 21, 2011, Row-lette offered to place Wall back on the participation list provided he could verify that Wall’s belongings had actually been lost. Rowlette then rescinded the offer once Wall stated that he intended to pursue a formal adjudication of the matter regardless. Taking this fact in the light most favorable to Wall, it indicates that the officials were more interested in protecting their earlier decision than in honestly discerning whether Wall should be permitted to participate.

In short, Wall has alleged that the defendants ignored numerous signs that he is a practicing Muslim who was merely seeking to exercise his genuinely held beliefs. By applying the policy in so rigid a manner, the restriction lost whatever “valid, rational connection” to the government’s stated interest that might have existed at the time it was adopted. 13

An analysis of the remaining Turner factors also supports the plaintiffs claims. *501 The second factor asks whether “alternative means of exercising the right ... remain open to [the] prison[er].” 482 U.S. at 90, 107 S.Ct. 2254. It is clear that Wall was absolutely precluded from observing Ramadan because of the defendants’ actions. When he attempted to adhere to the fasting requirements on his own by storing food in his cell, he was threatened with disciplinary action. The only alternative means proffered by the defendants is that had Wall obeyed the policy, he would have been allowed to participate. This suggestion, of course, it is not an “alternative means” at all, it is merely a reiteration of the same rigid requirements Wall was unable to meet. 14

We also believe that the third Turner factor, which examines the impact the requested accommodation would have on the prison’s efficient operation, also supports the plaintiffs claim. 482 U.S. at 92, 107 S.Ct. 2254. We are not satisfied that the defendants have sufficiently explained how a less restrictive policy would have imposed a significant burden on prison resources. The defendants contend generally that Ramadan is expensive because participants require special meals, and the schedules of both inmates and guards must be rearranged to accommodate pre-dawn and post-sunset meals. However, the record is void of any specific information regarding these purported costs, and we are not content to permit a prison to deny an inmate’s constitutional right in the face of such generalized concerns. This is especially so in light of the negligible costs associated with adding one additional inmate to an already existent program. Nor have the defendants presented a convincing argument why an individualized interpretation in Wall’s case would have been unduly burdensome. To the contrary, Wall presented the officials with significant evidence of his Muslim faith, which the defendants could have accepted without the need to conduct any further investigation on their own.

Finally, we are satisfied that there existed “easy[] [and] obvious alternatives” to the challenged regulation. Id. at 93, 107 S.Ct. 2254. This is most plainly seen in the fact that ROSP ultimately changed its policy, which has since allowed Wall and others to observe Ramadan without incident. Additionally, a VDOC guidance document issued June 25, 2010 addressed how inmates who were not on an institution’s designated religious pass list could demonstrate eligibility for Ramadan observance. One consideration utilized was past involvement in Ramadan fasting. ROSP, which does not maintain religious pass lists due to its status as a segregation facility, could have utilized the same, less restrictive criterion for determining eligibility.

In sum, viewing the current record in the light most favorable to the plaintiff, the defendants’ application of the 2010 Ramadan policy to Wall was unconstitutional. The defendants relied exclusively on a narrow set of parameters while ignoring obvious indications of the sincerity of Wall’s *502 beliefs. The First Amendment demands a more reasoned approach, even within the difficult confines of a prison environment. 15

2.

Having established a claim for a constitutional violation, we must now ask whether the defendants transgressed law that was “clearly established” at the time of the violation. Ridpath, 447 F.3d at 306. We conclude that, given Wall’s circumstances, his right to participate in Ramadan was clearly established, and the defendants are therefore not entitled to qualified immunity-

As noted, we have previously held that under “the Free Exercise Clause ... a prisoner has a clearly established ... right to a diet consistent with his ... religious scruples, including proper food during Ramadan.” Lovelace, 472 F.3d at 198-99 (emphasis added) (internal quotation marks and citation omitted). Further, “[a] prison official violates this clearly established right if he intentionally and without sufficient justification denies an inmate his religiously mandated diet.” Id. at 199. We take these statements to mean quite exactly what they say: that Wall’s right to participate in Ramadan was clearly established, and when the defendants abridged this right without first satisfying Turner’s reasonableness test, they subjected themselves to the potential for liability. As expressed above, the defendants’ application of their policy to Wall was unnecessarily strict. They overlooked (at best) significant evidence that Wall was, in fact, a practicing Muslim who was entitled to participate in Ramadan. We cannot conclude that a reasonable official in the defendants’ position, giving proper consideration to our statement in Lovelace that the right is clearly established, and to Turner’s objective reasonableness test, would have felt it permissible to apply the policy in so strict a fashion.

The defendants attempt to avoid this rather straightforward result by arguing that there is a lack of case law elucidating exactly how prisons may utilize sincerity tests in determining eligibility for religious accommodations. While it may be true that we have never specifically evaluated a sincerity test, 16 this argument overlooks the broader right at issue: that inmates are entitled to religious dietary accommodations absent a legitimate reason to the contrary. As we have previously stated, clearly established “includes not only already specifically adjudicated rights, but those manifestly included within more gen *503 eral applications of the core constitutional principle invoked.” Pritchett v. Alford, 973 F.2d 307, 314 (4th Cir.1992). In light of our unequivocal statement in Lovelace that inmates are entitled to religious dietary accommodations, we need not to have previously passed judgment on the appropriateness of particular sincerity tests in order to demand that prison officials act reasonably in administering that right. An expectation of reasonableness in this context is not a high bar, and does not punish officials for “bad guesses in gray areas.” Maciariello v. Sumner, 973 F.2d 295, 298 (4th Cir.1992). To the contrary, it offers only a minimal level of protection to inmates seeking to exercise their constitutionally protected rights.

Having found that the plaintiff has established a claim for a violation of his clearly established First Amendment rights, we vacate the district court’s grant of summary judgment on the plaintiffs First Amendment claim for damages.

III.

For the reasons explained above, we vacate the district court’s decision concluding that (1) the plaintiffs equitable claims are moot, and (2) the defendants are entitled to qualified immunity on the plaintiffs First Amendment damages claim. In so doing, we necessarily find that the plaintiffs claim under RLUIPA survives summary judgment as well, as such claims are evaluated under the same factors, but subject to a less demanding standard of proof. See Lovelace, 472 F.3d at 190. Accordingly, we remand this action to the district court for further proceedings as appropriate.

VACATED AND REMANDED.

Mr. Justice Stewart

delivered the opinion of the Court.

Section 706 (k) of Title VII of the Civil Rights Act of 1964 provides:

“In any action or proceeding under this title the court, *414 in its discretion, may allow the prevailing party ... a reasonable attorney’s fee . ...” 1

The question in this case is under what circumstances an attorney’s fee should be allowed when the defendant is the prevailing party in a Title VII action' — a question about which the federal courts have expressed divergent views.

I

Two years after Rosa Helm had filed a Title VII charge of racial discrimination against the petitioner Christiansburg Garment Co. (company), the Equal Employment Opportunity Commission notified her that its conciliation efforts had failed and that she had the right to sue the company in federal court. She did not do so. Almost two years later, in 1972, Congress enacted amendments to Title VII. 2 Section 14 of these amendments authorized the Commission to sue in its own name to prosecute “charges pending with the Commission” on the effective date of the amendments. Proceeding under this section, the Commission sued the company, alleging that it had engaged in unlawful employment practices in violation of the amended Act. The company moved for summary judgment on the ground, inter alia, that the Rosa Helm charge had not been “pending” before the Commission when the 1972 amendments took effect. The District Court agreed, and granted summary judgment in favor of the company. 376 F. Supp. 1067 (WD Va). 3

*415 The company then petitioned for the allowance of attorney’s fees against the Commission pursuant to § 706 (k) of Title VII. Finding that “the Commission’s action in bringing the suit cannot be characterized as unreasonable or meritless/’ the District Court concluded that “an award of attorney’s fees to petitioner is not justified in this case.” 4 A divided Court of Appeals affirmed, 550 F. 2d 949 (CA4), and we granted cer-tiorari to consider an important question of federal law, 432 U. S. 905.

II

It is the general rule in the United States that in the absence of legislation providing otherwise, litigants must pay their own attorney’s fees. Alyeska Pipeline Co. v. Wilderness Society, 421 U. S. 240. Congress has provided only limited exceptions to this rule “under selected statutes granting or protecting various federal rights.” Id., at 260. Some of these statutes make fee awards mandatory for prevailing plaintiffs; 5 others make awards permissive but limit them to certain parties, *416 usually prevailing plaintiffs. 6 But many of the statutes are more flexible, authorizing the award of attorney’s fees to either plaintiffs or defendants, and entrusting the effectuation of the statutory policy to the discretion of the district courts. 7 Section 706 (k) of Title VII of the Civil Nights Act of 1964 falls into this last category, providing as it does that a district court may in its discretion allow an attorney’s fee to the prevailing party.

In Newman v. Piggie Park Enterprises, 390 U. S. 400, the Court considered a substantially identical statute authorizing the award of attorney’s fees under Title II of the Civil Rights Act of 1964. 8 In that case the plaintiffs had prevailed, and the Court of Appeals had held that they should be awarded their attorney’s fees “only to- the extent that the respondents’ defenses had been advanced 'for purposes of delay and not in good faith.’ ” Id., at 401. We ruled that this “subjective standard” did not properly effectuate the purposes of the counsel-fee provision of Title II. Relying primarily on the intent of Congress to cast a Title II plaintiff in the role of “a 'private attorney general,’ vindicating a policy that Congress considered of the highest priority,” we held that a prevailing plaintiff under Title II “should ordinarily recover an attorney’s fee unless special circumstances would render such an award *417 unjust.” Id., at 402. We noted in passing that if the objective of Congress had been to permit the award of attorney’s fees only against defendants who h'ad acted in bad faith, “no new statutory provision would have been necessary,” since even the American common-law rule allows the award of attorney’s fees in those exceptional circumstances. Id., at 402 n. 4. 9

In Albemarle Paper Co. v. Moody, 422 U. S. 405, the Court made clear that the Piggie Park standard of awarding attorney’s fees to a successful plaintiff is equally applicable in an action under Title VII of the Civil Rights Act. 422 U. S., at 415. See also Northcross v. Memphis Board of Education, 412 U. S. 427, 428. It can thus be taken as established, as the parties in this case both acknowledge, that under § 706 (k) of Title VII a prevailing plaintiff ordinarily is to be awarded attorney’s fees in all but special circumstances. 10

Ill

The question in the case before us is what standard should inform a district court’s discretion in deciding whether to award attorney’s fees to a successful defendant in a Title VII action. Not surprisingly, the parties in addressing the question in their briefs and oral arguments have taken almost diametrically opposite positions. 11

The company contends that the Piggie Park criterion for a successful plaintiff should apply equally as a guide to the *418 award of attorney’s fees to a successful defendant. Its submission, in short, is that every prevailing defendant in a Title VII action should receive an allowance of attorney’s fees “unless special circumstances would render such an award unjust.” 12 The respondent Commission, by contrast, argues that the prevailing defendant should receive an award of attorney’s fees only when it is found that the plaintiff’s action was brought in bad faith. We have concluded that neither of these positions is correct.

A

Relying on what it terms “the plain meaning of the statute,” the company argues that the language of § 706 (k) admits of only one interpretation: “A prevailing defendant is entitled to an award of attorney’s fees on the same basis as a prevailing plaintiff.” But the permissive and discretionary language of the statute does not even invite, let alone require, such a mechanical construction. The terms of § 706 (k) provide no indication whatever of the circumstances under which either a plaintiff or a defendant should be entitled to' attorney’s fees. And a moment’s reflection reveals that there are at least two strong equitable considerations counseling an attorney’s fee award to a prevailing Title VII plaintiff that are wholly absent in the case of a prevailing Title VII defendant.

First, as emphasized so forcefully in Piggie Park, the plaintiff is the chosen instrument of Congress to vindicate “a policy that Congress considered of the highest priority.” 390 U. S., at 402. Second, when a district court awards counsel fees to a prevailing plaintiff, it is awarding them against a violator of federal law. As the Court of Appeals clearly perceived, “these policy considerations which support the award of fees to a *419 prevailing plaintiff are not present in the case of a prevailing defendant.” 550 F. 2d, at 951. A successful defendant seeking counsel fees under § 706 (k) must rely on quite different equitable considerations.

But if the company’s position is untenable, the Commission’s argument also misses the mark. It seems clear, in short, that in enacting § 706 (k) Congress did not intend to permit the award of attorney’s fees to a prevailing defendant only in a situation where the plaintiff was motivated by bad faith in bringing the action. As pointed out in Piggie Park, if that had been the intent of Congress, no statutory provision would have been necessary, for it has long been established that even under the American -common-law rule attorney’s fees may be awarded against a party who has proceeded in bad faith. 13

Furthermore, while it was certainly the policy of Congress that Title VII plaintiffs should vindicate “a policy that Congress considered of the highest priority,” Piggie Park, 390 U. S., at 402, it is equally certain that Congress entrusted the ultimate effectuation of that policy to the adversary judicial process, Occidental Life Ins. Co. v. EEOC, 432 U. S. 355. A fair adversary process presupposes both a vigorous prosecution and a vigorous defense. It cannot be lightly assumed that in enacting § 706 (k), Congress intended to' distort that process by giving the private plaintiff substantial incentives to sue, while foreclosing to the defendant the possibility of recovering his expenses in resisting even a groundless action unless he can show that it was brought in bad faith.

*420 B

The sparse legislative history of § 706 (k) reveals little more than the barest outlines of a proper accommodation of the competing considerations we have discussed. The only specific reference to § 706 (k) in the legislative debates indicates that the fee provision was included to “make it easier for a plaintiff of limited means to bring a meritorious suit.” 14 During the ¿Senate floor discussions of the almost identical attorney’s fee provision of Title II, however, several Senators explained that its allowance of awards to defendants would serve “to deter the bringing of lawsuits without foundation,” 15 “to discourage frivolous suits,” 16 and “to diminish the likelihood of unjustified suits being brought.” 17 If anything can be gleaned from these fragments of legislative history, it is that while Congress wanted to clear the way for suits to be brought under the Act, it also wanted to protect defendants from burdensome litigation having no legal or factual basis. The Court of Appeals for the District of Columbia Circuit seems to have drawn the maximum significance from the Senate debates when it concluded:

“[From these debates] two purposes for § 706 (k) emerge. First, Congress desired to 'make it easier for a plaintiff of limited means to bring a meritorious suit’.... But second, and equally important, Congress intended to 'deter the bringing of lawsuits without foundation’ by providing that the 'prevailing party’ — be it plaintiff or defendant — could obtain legal fees.” Grubbs v. Butz, 179 U. S. App. D. C. 18, 20, 648 F. 2d 973, 975.

The first federal appellate court to consider what criteria should govern the award of attorney’s fees to a prevailing *421 Title VII defendant was the Court of Appeals for the Third Circuit in United States Steel Corp. v. United States, 519 F. 2d 359. There a District Court had denied a fee award to a defendant that had successfully resisted a Commission demand for documents, the court finding that the Commission’s action had not been “ 'unfounded, meritless, frivolous or vexatiously brought.’ ” Id., at 363. The Court of Appeals concluded that the District Court had not abused its discretion in denying the award. Id., at 365. A similar standard was adopted by the Court of Appeals for the Second Circuit in Carrion v. Yeshiva University, 535 F. 2d 722. In upholding an attorney’s fee award to a successful defendant, that court stated that such awards should be permitted “not routinely, not simply because he succeeds, but only where the action brought is found to be unreasonable, frivolous, meritless or vexatious.” Id., at 727. 18

To the extent that abstract words can deal with concrete cases, we think that the concept embodied in the language adopted by these two Courts of Appeals is correct. We would qualify their words only by pointing out that the term “merit-less” is to be understood as meaning groundless or without foundation, rather than simply that the plaintiff has ultimately lost his case, and that the term “vexatious” in no way implies that the plaintiff’s subjective bad faith is a necessary prerequisite to a fee award against him. In sum, a district court may in its discretion award attorney’s fees to a prevailing defendant in a Title VII case upon a finding that the plaintiff’s action was frivolous, unreasonable, or without foundation, even though not brought in subjective bad faith.

In applying these criteria, it is important that a district court resist the understandable temptation to engage in post *422 hoc reasoning by concluding that, because a plaintiff did not ultimately prevail, his action must have been unreasonable or without foundation. This kind of hindsight logic could discourage all but the most airtight claims, for seldom can a prospective plaintiff be sure of ultimate success. No matter how honest one’s belief that he has been the victim of discrimination, no matter how meritorious one’s claim may appear at the outset, the course of litigation is rarely predictable. Decisive facts may not emerge until discovery or trial. The law may change or clarify in the midst of litigation. Even when the law or the facts appear questionable or unfavorable at the outset, a party may have an entirely reasonable ground for bringing suit.

That § 706 (k) allows fee awards only to prevailing private plaintiffs should assure that this statutory provision will not in itself operate as an incentive to the bringing of claims that have little chance of success. 19 To take the further step of assessing attorney’s fees against plaintiffs simply because they do not finally prevail would substantially add to the risks inhering in most litigation and would undercut the efforts of Congress to promote the vigorous enforcement of the provisions of Title VII. Hence, a plaintiff should not be assessed his opponent’s attorney’s fees unless a court finds that his claim was frivolous, unreasonable, or groundless, or that the plaintiff continued to litigate after it clearly became so. And, needless to say, if a plaintiff is found to have brought or continued such a claim in had faith, there will be an even stronger basis for charging him with the. attorney’s fees incurred by the defense. 20

*423 IV

In denying attorney’s fees to the company in this case, the District Court focused on the standards we have discussed. The court found that “the Commission’s action in bringing the suit cannot be characterized as unreasonable or meritless” because “the basis upon which petitioner prevailed was an *424 issue of first impression requiring judicial resolution” and because the “Commission’s statutory interpretation of § 14 of the 1972 amendments was not frivolous.” The court thus exercised its discretion squarely within the permissible bounds of § 706 (k). Accordingly, the judgment of the Court of Appeals upholding the decision of the District Court is affirmed.

It is so ordered.

*600CHIEF Justice Rehnquist

delivered the opinion of the Court.

Numerous federal statutes allow courts to award attorney’s fees and costs to the “prevailing party.” The question presented here is whether this term includes a party that has failed to secure a judgment on the merits or a court-ordered consent decree, but has nonetheless achieved the desired result because the lawsuit brought about a voluntary change in the defendant’s conduct. We hold that it does not.

Buekhannon Board and Care Home, Inc., which operates care homes that provide assisted living to their residents, failed an inspection by the West Virginia Office of the State Fire Marshal because some of the residents were incapable of “self-preservation” as defined under state law. See W. Va. Code §§16-5H-1, 16-5H-2 (1998) (requiring that all residents of residential board and care homes be capable of “self-preservation,” or capable of moving themselves “from situations involving imminent danger, such as fire”); W. Va. Code of State Rules, tit. 87, ser. 1, § 14.07(1) (1995) (same). On October 28, 1997, after receiving cease-and-desist orders requiring the closure of its residential care facilities within 30 days, Buekhannon Board and Care Home, Ine., on behalf of itself and other similarly situated homes and residents (hereinafter petitioners), brought suit in the United States *601District Court for the Northern District of West ’Virginia against the State of West Virginia, two of its agencies, and 18 individuals (hereinafter respondents), seeking declaratory and injunctive relief1 that the “self-preservation” requirement violated the Fair Housing Amendments Act of 1988 (FHAA), 102 Stat. 1619, 42 U. S. C. § 3601 et seq., and the Americans with Disabilities Act of 1990 (ADA), 104 Stat. 327, 42U.S.C. §12101 et seq.

Respondents agreed to stay enforcement of the cease-and-desist orders pending resolution of the case and the parties began discovery. In 1998, the West Virginia Legislature enacted two bills eliminating the “self-preservation” requirement, see S. 627,11998 W. Va. Aets 983-986 (amending regulations); H. R. 4200, II1998 W. Ya. Aets 1198-1199 (amending statute), and respondents moved to dismiss the ease as moot. The District Court granted the motion, finding that the 1998 legislation had eliminated the allegedly offensive provisions and that there was no indication that the West Virginia Legislature would repeal the amendments.2

Petitioners requested attorney’s fees as the “prevailing party” under the FHAA, 42 U.S.C. § 3613(c)(2) (“[T]he court, in its discretion, may allow the prevailing party... a reasonable attorney’s fee and costs”), and ADA, 42 U. S. C. § 12205 (“[T]he court... , in its discretion, may allow the prevailing party ... a reasonable attorney’s fee, including litigation expenses, and costs”). Petitioners argued that they were entitled to attorney’s fees under the “catalyst theory,” which posits that a plaintiff is a “prevailing party” if it achieves the desired result because the lawsuit brought about a voluntary change in the defendant’s conduct. Al*602though most Courts of Appeals recognize the "catalyst theory,”3 the Court of Appeals for the Fourth Circuit rejected it in S-1 and S-2 v. State Bd. of Ed. of N. C., 21 F. 3d 49, 51 (1994) (en bane) (“A person may not be a ‘‘prevailing party* . . . except by virtue of having obtained an enforceable judgment, consent decree, or settlement giving some of the legal relief sought”). The District Court accordingly denied the motion and, for the same reason, the Court of Appeals affirmed in an unpublished, per curiam opinion. Judgt. order reported at 208 F. 3d 819 (CA4 2000).

To resolve the disagreement amongst the Courts of Appeals, we granted certiorari, 530 U. S. 1304 (2000), and now affirm.

In the United States, parties are ordinarily required to bear their own attorney’s fees — the prevailing party is not entitled to collect from the loser. See Alyeska Pipeline Service Co. v. Wilderness Society, 421 U.S. 240, 247 (1975). Under this "American Rule,” we follow “a general practice of not awarding fees to a prevailing party absent explicit statutory authority.” Key Tronic Corp. v. United States, 511 U.S. 809, 819 (1994). Congress, however, has authorized the award of attorney’s fees to the “prevailing party” in numerous statutes in addition to those at issue here, such as the Civil Rights Act of 1964,78 Stat. 259,42 U. S. C. §2000e-5(k), the Voting Rights Act Amendments of 1975, 89 Stat. 402, 42 U. S. C. § 19732(e), and the Civil Rights Attorney’s *603Fees Awards Act of 1976, 90 Stat. 2641, 42 U. S. C. §1988. See generally Marek v. Chesny, 473 U.S. 1, 43-51 (1985) (Appendix to opinion of Brennan, J., dissenting).4

In designating those parties eligible for an award of litigation costs, Congress employed the term “prevailing party,” a legal term of art. Black’s Law Dictionary 1145 (7th ed. 1999) defines “prevailing party” as “[a] party in whose favor a judgment is rendered, regardless of the amount of damages awarded <in certain cases, the court will award attorney’s fees to the prevailing party>. —Also termed successful party” This view that a “prevailing party” is one who has been awarded some relief by the court can be distilled from our prior cases.5

In Hanrahan v. Hampton, 446 U.S. 754, 758 (1980) (per curiam), we reviewed the legislative history of § 1988 and found that “Congress intended to permit the interim award of counsel fees only when a party has prevailed on the merits of at least some of his claims.” Our “[rjespect for ordinary language requires that a plaintiff receive at least some relief on the merits of his claim before he can be said to prevail.” *604Hewitt v. Helms, 482 U. S. 755, 760 (1987). We have held that even an award of nominal damages suffices under this test. See Farrar v. Hobby, 506 U.S. 108 (1992).6

In addition to judgments on the merits, we have held that settlement agreements enforced through a consent decree may serve as the basis for an award of attorney’s fees. See Maher v. Gagne, 448 U.S. 122 (1980). Although a consent decree does not always include an admission of liability by the defendant, see, e. g., id., at 126, n. 8, it nonetheless is a court-ordered “ehang[e] [in] the legal relationship between [the plaintiff] and the defendant.” Texas State Teachers Assn. v. Garland Independent School Dist., 489 U.S. 782, 792 (1989) (citing Hewitt, supra, at 760-761, and Rhodes v. Stewart, 488 U. S. 1, 3-4 (1988) (per curiam)).7 These decisions, taken together, establish that enforceable judgments on the merits and court-ordered consent decrees create the “material alteration of the legal relationship of the parties” necessary to permit an award of attorney’s fees. 489 U.S., at 792-793; see also Hanrakan, supra, at 757 (“[I]t seems clearly to have been the intent of Congress to permit... an interlocutory award only to a party who has established his entitlement to some relief on the merits of his claims, either in the trial court or on appeal” (emphasis added)).

*605We think, however, the “catalyst theory” falls on the other side of the line from these examples. It allows an award where there is no judicially sanctioned change in the legal relationship of the parties. Even under a limited form of the “catalyst theory,” a plaintiff could recover attorney’s fees if it established that the “complaint had sufficient merit to withstand a motion to dismiss for lack of jurisdiction or failure to state a claim on which relief may be granted.” Brief for United States as Amicus Curiae 27. This is not the type of legal merit that our prior decisions, based upon plain language and congressional intent, have found necessary. Indeed, we held in Hewitt that an interlocutory ruling that reverses a dismissal for failure to state a claim “is not the stuff of which legal victories are made.” 482 U. S., at 760. See also Hanrahan, supra, at 754 (reversal of a directed verdict for defendant does not make plaintiff a “prevailing party”). A defendant’s voluntary change in conduct, although perhaps accomplishing what the plaintiff sought to achieve by the lawsuit, lacks the necessary judicial imprimatur on the change. Our precedents thus counsel against holding that the term “prevailing party” authorizes an award of attorney’s fees without a corresponding alteration in the legal relationship of the parties.

The dissenters chide us for upsetting “long-prevailing Circuit precedent.” Post, at 622 (opinion of Ginsbukg, J.) (emphasis added). But, as Justice Scalia points out in his concurrence, several Courts of Appeals have relied upon dicta in our prior eases in approving the “catalyst theory.” See post, at 621-622; see also supra, at 608, n. 5. Now that the issue is squarely presented, it behooves us to reconcile the plain language of the statutes with our prior holdings. We have only awarded attorney’s fees where the plaintiff has received a judgment on the merits, see, e. g., Farrar, supra, at 112, or obtained a court-ordered consent decree, Maher, supra, at 129-130 — we have not awarded attorney’s fees where the plaintiff has secured the reversal of a directed *606verdict, see Hanrahan, 446 U.S., at 759, or acquired a judicial pronouncement that the defendant has violated the Constitution unaccompanied by “judicial relief,” Hewitt, supra, at 760 (emphasis added). Never have we awarded attorney’s fees for a nonjudieial “alteration of actual circumstances.” Post, at 638 (dissenting opinion). While urging an expansion of our precedents on this front, the dissenters would simultaneously abrogate the “merit” requirement of our prior cases and award attorney’s fees where the plaintiff’s claim “was at least colorable” and “not... groundless.” Post, at 627 (internal quotation marks and citation omitted). We cannot agree that the term “prevailing party” authorizes federal courts to award attorney’s fees to a plaintiff who, by simply filing a nonfrivolous but nonetheless potentially meritless lawsuit (it will never be determined), has reached the “sought-after destination” without obtaining any judicial relief. Post, at 634 (internal quotation marks and citation omitted).8

*607Petitioners nonetheless argue that the legislative history of the Civil Rights Attorney's Fees Awards Act supports a broad reading of “prevailing party" which includes the “catalyst theory.” We doubt that legislative history could overcome what we think is the rather clear meaning of “prevailing party” — the term actually used in the statute. Since we resorted to such history in Garland, 489 U.S., at 790, Maher, 448 U. S., at 129, and Hanrakan, supra, at 756-757, however, we do likewise here.

The House Report to § 1988 states that “[t]he phrase ‘prevailing party' is not intended to be limited to the victor only after entry of a final judgment following a full trial on the merits,” H. R. Rep. No. 94-1558, p. 7 (1976), while the Senate Report explains that “parties may be considered to have prevailed when they vindicate rights through a consent judgment or without formally obtaining relief,” S. Rep. No. 94-1011, p. 5 (1976). Petitioners argue that these Reports and their reference to a 1970 decision from the Court of Appeals for the Eighth Circuit, Parham v. Southwestern Bell Telephone Co., 433 F. 2d 421 (1970), indicate Congress' intent to adopt the “catalyst theory.”9 We think the legislative his*608tory cited by petitioners is at best ambiguous as to the availability of the “catalyst theory” for awarding attorney’s fees. Particularly in view of the “American Rule” that attorney’s fees will not be awarded absent “explicit statutory authority,” such legislative history is clearly insufficient to alter the accepted meaning of the statutory term. Key Tronic, 511 U.S., at 819; see also Hanrahan, supra, at 758 (“[0]nly when a party has prevailed on the merits of at least some of his claims ... has there been a determination of the ‘substantial rights of the parties,’ which Congress determined was a necessary foundation for departing from the usual rule in this country that each party is to bear the expense of his own attorney” (quoting H. R. Rep. No. 94-1558, at 8)).

Petitioners finally assert that the “catalyst theory” is necessary to prevent defendants from unilaterally mooting an action before judgment in an effort to avoid an award of attorney’s fees. They also claim that the rejection of the “catalyst theory” will deter plaintiffs with meritorious but expensive eases from bringing suit. We are skeptical of these assertions, which are entirely speculative and unsupported by any empirical evidence (e. g., whether the number of suits brought in the Fourth Circuit has declined, in relation to other Circuits, since the decision in S-l and S-2).

Petitioners discount the disincentive that the “catalyst theory” may have upon a defendant’s decision to voluntarily change its conduct, conduct that may not be illegal. “The defendants’ potential liability for fees in this kind of litigation can be as significant as, and sometimes even more significant than, their potential liability on the merits,” Evans v. Jeff D., 475 U.S. 717, 734 (1986), and the possibility of being assessed attorney’s fees may well deter a defendant from altering its conduct.

And petitioners’ fear of mischievous defendants only materializes in claims for equitable relief, for so long as the *609plaintiff has a cause of action for damages, a defendant’s change in conduct -will not moot the case.10 Even then, it is not clear how often courts will find a case mooted: “It is well settled that a defendant’s voluntary cessation of a challenged practice does not deprive a federal court of its power to determine the legality of the practice” unless it is “absolutely clear that the allegedly wrongful behavior could not reasonably be expected to recur.” Friends of Earth, Inc. v. Laidlaw Environmental Services (TOG), Inc., 528 U.S. 167, 189 (2000) (internal quotation marks and citations omitted). If a case is not found to be moot, and the plaintiff later procures an enforceable judgment, the court may of course award attorney’s fees. Given this possibility, a defendant has a strong incentive to enter a settlement agreement, where it can negotiate attorney’s fees and costs. Cf. Marek v. Chesny, 473 U.S., at 7 (“[M]any a defendant would be unwilling to make a binding settlement offer on terms that left it exposed to liability for attorney’s fees in whatever amount the court might fix on motion of the plaintiff” (internal quotation marks and citation omitted)).

We have also stated that “[a] request for attorney’s fees should not result in a second major litigation,” Hensley v. Eckerhart, 461 U. S. 424, 437 (1983), and have accordingly avoided an interpretation of the fee-shifting statutes that would have “spawnfed] a second litigation of significant dimension,” Garland, supra, at 791. Among other things, a “catalyst theory” hearing would require analysis of the defendant’s subjective motivations in changing its conduct, an analysis that “will likely depend on a highly faetbound inquiry and may turn on reasonable inferences from the nature and timing of the defendant’s change in conduct.” *610Brief for United States as Amicus Curiae 28. Although we do not doubt the ability of district courts to perform the nuaneed “three thresholds” test required by the “catalyst theory” — whether the claim was colorable rather than groundless; whether the lawsuit was a substantial rather than an insubstantial cause of the defendant's change in conduct; whether the defendant’s change in conduct was motivated by the plaintiff’s threat of victory rather than threat of expense, see post, at 627-628 (dissenting opinion)— it is clearly not a formula for “ready administrability.” Burlington v. Dague, 505 U.S. 557, 566 (1992).

Given the clear meaning of “prevailing party” in the fee-shifting statutes, we need not determine which way these various policy arguments cut. In Alyeska, 421 U.S., at 260, we said that Congress had not “extended any roving authority to the Judiciary to allow counsel fees as costs or otherwise whenever the courts might deem them warranted.” To disregard the clear legislative language and the holdings of our prior cases on the basis of such policy arguments would be a similar assumption of a “roving authority.” For the reasons stated above, we hold that the “catalyst theory” is not a permissible basis for the award of attorney’s fees under the FHAA, 42 U. S. C. § 3613(e)(2), and ADA, 42 U.S.C. §12205.

The judgment of the Court of Appeals is

Affirmed.

Justice Scalia,

concurring.

I join the opinion of the Court in its entirety, and write to respond at greater length to the contentions of the dissent.

I

“Prevailing party” is not some newfangled legal term invented for use in late-20th-eentury fee-shifting statutes. *611“[B]y the long established practice and universally recognized rule of the common law, in actions at law, the prevailing party is entitled to recover a judgment for costs . . . .” Mansfield, C. & L. M. R. Co. v. Swan, 111 U.S. 379, 387 (1884).

“Costs have usually been allowed to the prevailing party, as incident to the judgment, since the statute 6 Edw. I, c. 1, §2, and the same rule was acknowledged in the courts of the States, at the time the judicial system of the United States was organized....

“Weighed in the light of these several provisions in the Judiciary Act [of 1789], the conclusion appears to be clear that Congress intended to allow costs to the prevailing party, as incident to the judgment. . . .” The Baltimore, 8 Wall. 377, 388, 390 (1869).

The term has been found within the United States Statutes at Large since at least the Bankruptcy Act of 1867, which provided that “[t]he party prevailing in the suit shall be entitled to costs against the adverse party.” Act of Mar. 2,1867, eh. 176, §24, 14 Stat. 528. See also Act of Mar. 3, 1887, ch. 359, § 15,24 Stat. 508 (“If the Government of the United States shall put in issue the right of the plaintiff to recover the court may, in its discretion, allow costs to the prevailing party from the time of joining such issue”). A computer search shows that the term “prevailing party” appears at least 70 times in the current United States Code; it is no stranger to the law.

At the time 42 U. S. C. § 1988 was enacted, I know of no case, state or federal, in which — either under a statutory invocation of “prevailing party” or under the common-law rule — the “catalyst theory” was enunciated as the basis for awarding costs. Indeed, the dissent cites only one case in which (although the “catalyst theory” was not expressed) *612costs were awarded for a reason that the eatalyst theory would support, but today’s holding of the Court would not: Baldwin v. Chesapeake & Potomac Tel. Co., 156 Md. 552, 557, 144 A. 703, 705 (1929), where costs were awarded because “the granting of [appellee’s] motion to dismiss the appeal has made it unnecessary to inquire into the merits of the suit, and the dismissal is based on an act of appellee performed after both the institution of the suit and the entry of the appeal.” And that ease is irrelevant to the meaning of “prevailing party,” because it was a case in equity. While, as Mansfield observed, costs were awarded in actions at law to the “prevailing party,” see 111 U.S., at 387, an equity court could award costs “as the equities of the case might require,” Getz v. Johnston, 145 Md. 426, 433, 125 A. 689, 691 (1924). See also Born v. Bohn, 96 Md. 8,12-13, 53 A. 576, 577 (1902) (“The question of costs in equity eases is a matter resting in the sound discretion of the Court, from the exercise of which no appeal will lie” (internal quotation marks and citation omitted)).1 The other state or state-law eases the dis*613sent cites as awarding costs despite the absence of a judgment all involve a judicial finding — or its equivalent, an acknowledgment by the defendant — of the merits of plaintiff’s case.2 Moreover, the dissent cites not a single case in *614which this Court — or even any other federal court applying federal law prior to enactment of the fee-shifting statutes at issue here — has regarded as the “prevailing party” a litigant who left the courthouse emptyhanded. If the term means what the dissent contends, that is a remarkable absence of authority.

That a judicial finding of liability was an understood requirement of “prevailing” is confirmed by many statutes that use the phrase in a context that presumes the existence of a judicial ruling. See, e. g., 5 U. S. C. § 1221(g)(2) (“[i]f an employee ... is the prevailing party . . . and the decision is based on a finding of a prohibited personnel practice”); § 1221(g)(3) (providing for an award of attorney’s fees to the “prevailing party,” “regardless of the basis of the decision”); § 7701(b)(2)(A) (allowing the prevailing party to obtain an interlocutory award of the “relief provided in the decision”); 8 U. S. C. § 1324b(h) (permitting the administrative law judge to award an attorney’s fee to the prevailing party “if the losing party’s argument is without reasonable foundation in law and fact”); 18 ü. S. C. § 1864(e) (1994 ed., Supp. V) (allowing the district court to award the prevailing party its attorney’s fee “in addition to monetary damages”).

The dissent points out, post, at 629, that the Prison Litigation Reform Act of 1995 limits attorney’s fees to an amount “proportionately related to the court ordered relief for the violation.” This shows that sometimes Congress does explicitly “tightly bind fees to judgments,” ibid., inviting (the dissent believes) the conclusion that “prevailing party” does not fasten fees to judgments. That conclusion does not follow from the premise. What this statutory provision demonstrates, at most, is that use of the phrase “prevailing party” is not the only way to impose a requirement of court-ordered relief. That is assuredly true. But it would be no *615more rational to reject the normal meaning of "prevailing party” because some statutes produce the same result with different language, than it would be to conclude that, since there are many synonyms for the word “jump,” the word “jump” must mean something else.

It is undoubtedly true, as the dissent points out by quoting a nonlegal dictionary, see post, at 633-634, that the word “prevailing” can have other meanings in other contexts: “prevailing winds” are the winds that predominate, and the “prevailing party” in an election is the party that wins the election. But when “prevailing party” is used by courts or legislatures in the context of a lawsuit, it is a term of art. It has traditionally — and to my knowledge, prior to enactment of the first of the statutes at issue here, invariably— meant the party that wins the suit or obtains a finding (or an admission) of liability. Not the party that ultimately gets his way because his adversary dies before the suit comes to judgment; not the party that gets his way because circumstances so change that a victory on the legal point for the other side turns out to be a practical victory for him; and not the party that gets his way because the other side ceases (for whatever reason) its offensive conduct. If a nuisance suit is mooted because the defendant asphalt plant has gone bankrupt and ceased operations, one would not normally call the plaintiff the prevailing party. And it would make no difference, as far as the propriety of that characterization is concerned, if the plant did not go bankrupt but moved to a new location to avoid the expense of litigation. In one sense the plaintiff would have “prevailed”; but he would not be the prevailing party in the lawsuit. Words that have acquired a specialized meaning in the legal context must be accorded their legal meaning.

“[Wlhere Congress borrows terms of art in which are accumulated the legal tradition and meaning of centuries of practice, it presumably knows and adopts the cluster of ideas that were attached to each borrowed word in *616the body of learning from which it was taken and the meaning its use will convey to the judicial mind unless otherwise instructed. In such ease, absence of contrary direction may be taken as satisfaction with widely accepted definitions, not as a departure from them.” Morissette v. United States, 342 U.S. 246, 263 (1952).

The cases cited by the dissent in which we have “not treated Black’s Law Dictionary as preclusively definitive,” post, at 628-629, are inapposite. In both Pioneer Investment Services Co. v. Brunswick Associates Ltd. Partnership, 507 U.S. 380 (1993), and United States v. Rodgers, 466 U.S. 475 (1984), we rejected Black’s definition because it conflicted with our precedent. See Pioneer, supra, at 395-396, n. 14; Rodgers, supra, at 480. We did not, as the dissent would do here, simply reject a relevant definition of a word tailored to judicial settings in favor of a more general definition from another dictionary.

II

The dissent distorts the term “prevailing party” beyond its normal meaning for policy reasons, but even those seem to me misguided. They rest upon the presumption that the catalyst theory applies when “the suit’s merit led the defendant to abandon the fray, to switch rather than fight on, to accord plaintiff sooner rather than later the principal redress sought in the complaint,” post, at 622 (emphasis added). As the dissent would have it, by giving the term its normal meaning the Court today approves the practice of denying attorney’s fees to a plaintiff with a proven claim of discrimination, simply because the very merit of his claim led the defendant to capitulate before judgment. That is not the ease. To the contrary, the Court approves the result in Parham v. Southwestern Bell Tel. Co., 433 F. 2d 421 (CA8 1970), where attorney’s fees were awarded “after [a] finding that the defendant had acted unlawfully,” ante, at 607-*617608, and n. 9.3 What the dissent’s stretching of the term produces is something more, and something far less reasonable: an award of attorney’s fees when the merits of the plaintiff’s case remain unresolved — when, for all one knows, the defendant only “abandoned] the fray” because the cost of litigation — either financial or in terms of public relations— would be too great. In such a case, the plaintiff may have “prevailed” as Webster’s defines that term — “gain[ed] victory by virtue of strength or superiority,” see post, at 633. But I doubt it was greater strength in financial resources, or superiority in media manipulation, rather than superiority in legal merit, that Congress intended to reward.

*618It could be argued, perhaps, that insofar as abstract justice is concerned, there is little to choose between the dissent’s outcome and the Court’s: If the former sometimes rewards the plaintiff with a phony claim (there is no way of knowing), the latter sometimes denies fees to the plaintiff with a solid case whose adversary slinks away on the eve of judgment. But it seems to me the evil of the former far outweighs the evil of the latter. There is all the difference in the world between a rule that denies the extraordinary boon of attorney’s fees to some plaintiffs who are no less “deserving” of them than others who receive them, and a rule that causes the law to be the very instrument of wrong — exacting the payment of attorney’s fees to the extortionist.

It is true that monetary settlements and consent decrees can be extorted as well, and we have approved the award of attorney’s fees in cases resolved through such mechanisms. See ante, at 604 (citing eases). Our decision that the statute makes plaintiff a “prevailing party” under such circumstances was based entirely on language in a House Report, see Maher v. Gagne, 448 U.S. 122, 129 (1980), and if this issue were to arise for the first time today, I doubt whether I would agree with that result. See Hewitt v. Helms, 482 U.S. 755, 760 (1987) (Scalia, J.) (opining that “[rjespeet for ordinary language requires that a plaintiff receive at least some relief on the merits of his claim before he can be said to prevail” (emphasis added)). But in the case of court-approved settlements and consent decrees, even if there has been no judicial determination of the merits, the outcome is at least the product of, and bears the sanction of, judicial action in the lawsuit There is at least some basis for saying that the party favored by the settlement or decree prevailed in the suit. Extending the holding of Maher to a case in which no judicial action whatever has been taken stretches the term "prevailing party” (and the potential injustice that Maher produces) beyond what the normal mean*619ing of that term in the litigation context can conceivably support.

The dissent points out that petitioners’ object in bringing their suit was pot to obtain "a judge’s approbation,” but to “stop enforcement of a [West "Virginia] rule,” post, at 684; see also Hewitt, supra, at 761. True enough. But not even the dissent claims that if a petitioner accumulated attorney’s fees in preparing a threatened complaint, but never filed it prior to the defendant’s voluntary cessation of its offending behavior, the wannabe-but-never-was plaintiff could recover fees; that would be countertextual, since the fee-shifting statutes require that there be an “action” or “proceeding,” see 42 U. S. G. §§ 3613(d), 1988(b) (1994 ed., Supp. V) — which in legal parlance (though not in more general usage) means a lawsuit See post, at 643 (concluding that a party should be deemed prevailing as a result of a “postcomplaint payment or change in conduct” (emphasis added)). Does that not leave achievement of the broad congressional purpose identified by the dissent just as unsatisfactorily incomplete as the failure to award fees when there is no decree? Just as the dissent rhetorically asks why (never mind the language of the statute) Congress would want to award fees when there is a judgment, but deny fees when the defendant capitulates on the eve of judgment; so also it is fair for us to ask why Congress would want to award fees when suit has been filed, but deny fees when the about-to-be defendant capitulates under the threat of filing. Surely, it cannot be because determination of whether suit was actually contemplated and threatened is too difficult. All the proof takes is a threatening letter and a batch of timesheets. Surely that obstacle would not deter the Congress that (according to the dissent) was willing to let district judges pursue that much more evasive will-o’-the-wisp called “catalyst.” (Is this not why we have district courts?, asks the dissent, post, at 639-640.) My point is not that it would take no more twisting *620of language to produce prelitigation attorney’s fees than to produce the decreeless attorney’s fees that the dissent favors (though that may well be true). My point is that the departure from normal usage that the dissent favors cannot be justified on the ground that it establishes a regime of logical evenhandedness. There must be a cutoff of seemingly equivalent entitlements to fees — either the failure to file suit in time or the failure to obtain a judgment in time. The term “prevailing party” suggests the latter rather than the former. One does not prevail in a suit that is never determined.

The dissent’s ultimate worry is that today’s opinion will “impede access to court for the less well-heeled,” post, at 623. But, of course, the catalyst theory also harms the “less well-heeled,” putting pressure on them to avoid the risk of massive fees by abandoning a solidly defensible case early in litigation. Since the fee-shifting statutes at issue here allow defendants as well as plaintiffs to receive a fee award, we know that Congress did not intend to maximize the quantity of “the enforcement of federal law by private attorneys general,” ibid. Rather, Congress desired an appropriate level of enforcement — which is more likely to be produced by limiting fee awards to plaintiffs who prevail “on the merits,” or at least to those who achieve an enforceable “alteration of the legal relationship of the parties,” than by permitting the open-ended inquiry approved by the dissent.4

*621hH < H — I

The dissent points out that the catalyst theory has been accepted by “the clear majority of Federal Circuits,” ibid. But our disagreeing with a “clear majority” of the Circuits is not at all a rare phenomenon. Indeed, our opinions sometimes contradict the unanimous and longstanding interpretation of lower federal courts. See, e. g., McNally v. United States, 483 U.S. 350, 365 (1987) (Stevens, J., dissenting) (the Court’s decision contradicted “[e]very court to consider” the question).

The dissent’s insistence that we defer to the “clear majority” of Circuit opinion is particularly peculiar in the present case, since that majority has been nurtured and preserved by our own misleading dicta (to which I, unfortunately, contributed). Most of the Court of Appeals cases cited by the dissent, post, at 627, and n. 5, as reaffirming the catalyst theory after our decision in Farrar v. Hobby, 506 U.S. 103 (1992), relied on our earlier opinion in Hewitt. See Marbley v. Bane, 57 F. 3d 224, 234 (CA2 1995) (relying on Hewitt to support catalyst theory); Payne v. Board of Ed., 88 F. 3d 392, 397 (CA6 1996) (same); Baumgartner v. Harrisburg Housing Auth., 21 F. 3d 541, 548 (CA3 1994) (explicitly rejecting Farrar in favor of Hewitt); Zinn v. Shdlala, 35 F. 3d 273, 274-276 (CA7 1994) (same); Beard v. Teska, 31 F. 3d 942, 950-952 (CA10 1994) (same); Morris v. West Palm Beach, 194 F. 3d 1203, 1207 (CA11 1999) (same). Deferring to our colleagues’ own error is bad enough; but enshrining the error that we ourselves have improvidently suggested and blaming it on the near-unanimous judgment of our colleagues would surely be unworthy.5 Informing the Courts of Ap*622peals that our ill-considered dicta have misled them displays, it seems to me, not “disrespect,” but a most becoming (and well-deserved) humility.

* * *

The Court today concludes that a party cannot be deemed to have prevailed, for purposes of fee-shifting statutes such as 42 U. S. C. §§1988, 3618(c)(2) (1994 ed. and Supp. V), unless there has been an enforceable “alteration of the legal relationship of the parties.” That is the normal meaning of “prevailing party” in litigation, and there is no proper basis for departing from that normal meaning. Congress is free, of course, to revise these provisions — -but it is my guess that if it does so it will not create the sort of inequity that the catalyst theory invites, but will require the court to determine that there was at least a substantial likelihood that the party requesting fees would have prevailed.

Justice Ginsburg,

dissenting.

The Court today holds that a plaintiff whose suit prompts the precise relief she seeks does not “prevail,” and hence cannot obtain an award of attorney’s fees, unless she also secures a court entry memorializing her victory. The entry need not be a judgment on the merits. Nor need there be any finding of wrongdoing. A court-approved settlement will do.

The Court’s insistence that there be a document filed in court — a litigated judgment or court-endorsed settlement— upsets long-prevailing Circuit precedent applicable to scores of federal fee-shifting statutes. The decision allows a defendant to escape a statutory obligation to pay a plaintiff’s counsel fees, even though the suit’s merit led the defendant to abandon the fray, to switch rather than fight on, to accord plaintiff sooner rather than later the principal redress sought in the complaint. Concomitantly, the Court’s constricted *623definition of “prevailing party,” and consequent rejection of the “catalyst theory,” impede access to court for the less well heeled, and shrink the incentive Congress created for the enforcement of federal law by private attorneys general.

In my view, the “catalyst rule,” as applied by the clear majority of Federal Circuits, is a key component of the fee-shifting statutes Congress adopted to advance enforcement of civil rights. Nothing in history, precedent, or plain English warrants the anemic construction of the term “prevailing party” the Court today imposes.

I

Petitioner Buckhannon Board and Care Home, Inc. (Buck-hannon), operates residential care homes for elderly persons who need assisted living, but not nursing services. Among Buekhannon’s residents in October 1996 was 102-year-old Dorsey Pierce. Pierce had resided at Buckhannon for some four years. Her daughter lived nearby, and the care provided at Buckhannon met Pierce’s needs. Until 1998, West Virginia had a “self-preservation” rule prohibiting homes like Buckhannon from accommodating persons unable to exit the premises without assistance in the event of a fire. Pierce and two other Buckhannon residents could not get to a fire exit without aid. Informed of these residents’ limitations, West Virginia officials proceeded against Buckhannon for noncompliance with the self-preservation rule. On October 18,1996, three orders issued, each commanding Buckhan-non to “cease operating . . . and to effect relocation of [its] existing population within thirty (30) days.” App. 46-53.

Ten days later, Buckhannon and Pierce, together with an organization of residential homes and another Buckhannon resident (hereinafter plaintiffs), commenced litigation in Federal District Court to overturn the cease-and-desist orders and the self-preservation rule on which they rested. They sued the State, state agencies, and 18 officials (hereinafter defendants) alleging that the rule discriminated *624against persons with disabilities in violation of the Fair Housing Amendments Act of 1988 (FHAA), 42 U. S. C. § 3601 et seq., and the Americans with Disabilities Act of 1990 (ADA), 42 U. S. C. § 12101 et seq. Plaintiffs sought an immediate order stopping defendants from closing Buckhannon’s facilities, injunctive relief permanently barring enforcement of the self-preservation requirement, damages, and attorney’s fees.

On November 1, 1996, at a hearing on plaintiffs’ request for a temporary restraining order, defendants agreed to the entry of an interim order allowing Buekhannon to remain open without changing the individual plaintiffs’ housing and care. Discovery followed. On January 2, 1998, facing the state defendants’ sovereign immunity pleas, plaintiffs stipulated to dismissal of their demands for damages. In February 1998, in response to defendants’ motion to dispose of the remainder of the ease summarily, the District Court determined that plaintiffs had presented triable claims under the FHAA and ADA.

Less than a month after the District Court found that plaintiffs were entitled to a trial, the West Virginia Legislature repealed the self-preservation rule. Plaintiffs still allege, and seek to prove, that their suit triggered the statutory repeal. After the rule’s demise, defendants moved to dismiss the case as moot, and plaintiffs sought attorney’s fees as “prevailing parties” under the FHAA, 42 U. S. C. § 3613(c)(2), and the ADA, 42 U. S. C. § 12205.1

*625Finding no likelihood that West Virginia would reenact the self-preservation rule, the District Court agreed that the State’s action had rendered the case moot. Turning to plaintiffs’ application for attorney’s fees, the District Court followed Fourth Circuit precedent requiring the. denial of fees unless termination of the action was accompanied by a judgment, consent decree, or settlement.2 Plaintiffs did not appeal the mootness determination, and the Fourth Circuit affirmed the denial of attorney’s fees. In sum, plaintiffs were denied fees not because they failed to achieve the relief they sought. On the contrary, they gained the very change they sought through their lawsuit when West Virginia repealed the self-preservation rule that would have stopped Buckhannon from earing for people like Dorsey Pierce.3

Prior to 1994, every Federal Court of Appeals (except the Federal Circuit, which had not addressed the issue) concluded that plaintiffs in situations like Buekhannon’s and *626Pierce's could obtain a fee award if their suit acted as a “catalyst” for the change they sought, even if they did not obtain a judgment or consent decree.4 The Courts of Appeals found it “clear that a party may be considered to have prevailed even when the legal action stops short of final . . . judgment due to . . . intervening mootness.” Grano v. Barry, 783 F. 2d 1104, 1108 (CADC 1986). Interpreting the term “prevailing party” in “a practical sense,” Stewart v. Hannon, 675 F. 2d 846, 851 (CA7 1982) (citation omitted), federal courts across the country held that a party “prevails” for fee-shifting purposes when “its ends are accomplished as a result of the litigation,” Associated Builders & Contractors v. Orleans Parish School Bd., 919 F. 2d 374, 378 (CA5 1990) (citation and internal quotation marks omitted).

In 1994, the Fourth Circuit en bane, dividing 6-to-5, broke ranks with its sister courts. The court declared that, in light of Farrar v. Hobby, 506 U.S. 103 (1992), a plaintiff could *627not become a ‘‘prevailing party” without “an enforceable judgment, consent decree, or settlement.” S-l and S-2 v. State Bd. of Ed. of N. C, 21 F. 3d 49, 51 (1994). As the Court today acknowledges, see ante, at 603, n. 5, and as we have previously observed, the language on which the Fourth Circuit relied was dictum: Farrar “involved no catalytic effect”; the issue plainly “was not presented for this Court’s decision in FarrarFriends of Earth, Inc. v. Laidlaw Environmental Services (TOC), Inc., 528 U.S. 167, 194 (2000).

After the Fourth Circuit’s en banc ruling, nine Courts of Appeals reaffirmed their own consistently held interpretation of the term “prevail.”5 On this predominant view, “[sjecuring an enforceable decree or agreement may evidence prevailing party status, but the judgment or agreement simply embodies and enforces what is sought in bringing the lawsuit.... Victory can be achieved well short of a final judgment (or its equivalent) . . . . ” Marbley v. Bane, 57 F. 3d 224, 234 (CA2 1995) (Jacobs, J.).

The array of federal-court decisions applying the catalyst rule suggested three conditions necessary to a party’s qualification as “prevailing” short of a favorable final judgment or consent decree. A plaintiff first had to show that the defendant provided “some of the benefit sought” by the lawsuit. Wheeler v. Towanda Area School Dist., 950 F. 2d 128, 131 (CA31991). Under most Circuits’ precedents, a plaintiff had to demonstrate as well that the suit stated a genuine claim, i. e., one that was at least “colorable,” not “frivolous, unreasonable, or groundless.” Grano, 783 F. 2d, at 1110 (internal *628quotation marks and citation omitted). Plaintiff finally had to establish that her suit was a “substantial” or “significant” cause of defendant’s action providing relief. Williams v. Leatherbury, 672 F. 2d 549, 551 (GA5 1982). In some Circuits, to make this causation showing, plaintiff had to satisfy the trial court that the suit achieved results “by threat of victory,” not “by dint of nuisance and threat of expense.” Marbley, 57 F. 3d, at 234-235; see also Hooper v. Demco, Inc., 37 F. 3d 287, 293 (CA7 1994) (to render plaintiff “prevailing party,” suit “must have prompted the defendant... to act or cease its behavior based on the strength of the ease, not 'wholly gratuitously'”). One who crossed these three thresholds would be recognized as a “prevailing party” to whom the district court, “in its discretion,” supra, at 624-625, n. 1, could award attorney’s fees.

Developed over decades and in legions of federal-court decisions, the catalyst rule and these implementing standards deserve this Court’s respect and approbation.

II

A

The Court today detects a “clear meaning” of the term prevailing party, ante, at 610, that has heretofore eluded the large majority of courts construing those words. “Prevailing party,” today’s opinion announces, means “one who has been awarded some relief by the court,” ante, at 603. The Court derives this “clear meaning” principally from Black’s Law Dictionary, which defines a “prevailing party,” in critical part, as one “in whose favor a judgment is rendered,” ibid, (quoting Black’s Law Dictionary 1145 (7th ed. 1999)).

One can entirely agree with Black’s Law Dictionary that a party “in whose favor a judgment is rendered” prevails, and at the same time resist, as most Courts of Appeals have, any implication that only such a party may prevail. In prior cases, we have not treated Black’s Law Dictionary as preelu-*629sively definitive; instead, we have accorded statutory terms, including legal “termfs] of art,” ante, at 603 (opinion of the Court); ante, at 616 (Scalia, J., concurring), a contextual reading. See, e.g., Pioneer Investment Services Co. v. Brunswick Associates Ltd. Partnership, 507 U.S. 380, 395-396, n. 14 (1993) (defining “excusable negleet,” as used in Federal Rule of Bankruptcy Procedure 9006(b)(1), more broadly than Black’s defines that term); United States v. Rodgers, 466 U. S. 475, 479-480 (1984) (adopting “natural, nontechnical” definition of word “jurisdiction,” as that term is used in 18 U. S. C. § 1001, and declining to confine definition to “narrower, more technical meanings,” citing Black’s). Notably, this Court did not refer to Blaek’s Law Dictionary in Maher v. Gagne, 448 U.S. 122 (1980), which held that a consent decree could qualify a plaintiff as “prevailing.” The Court explained:

“The fact that [plaintiff] prevailed through a settlement rather than through litigation does not weaken her claim to fees. Nothing in the language of [42 U. S. C.] § 1988 conditions the District Court’s power to award fees on full litigation of the issues or on a judicial determination that the plaintiff’s rights have been violated.” Id., at 129.

The spare “prevailing party” language of the fee-shifting provision applicable in Maher, and the similar wording of the fee-shifting provisions now before the Court, contrast with prescriptions that so tightly bind fees to judgments as to exclude the application of a catalyst concept. The Prison Litigation Reform Act of 1995, for example, directs that fee awards to prisoners under § 1988 be “proportionately related to the court ordered relief for the violation.” 110 Stat. 1321-72, as amended, 42 U. S. C. § 1997e(d)(l)(B)(i) (1994 ed., Supp. V) (emphasis added). That statute, by its express terms, forecloses an award to a prisoner on a catalyst theory. But the FHAA and ADA fee-shifting prescriptions, modeled *630on 42 U. S. C. § 1988 unmodified, see supra, at 624-625, n. 1, do not similarly staple fee awards to “court ordered relief.” Their very terms do not foreclose a catalyst theory.

B

It is altogether true, as the concurring opinion points out, ante, at 610-611, that litigation costs other than attorney’s fees traditionally have been allowed to the “prevailing party,” and that a judgment winner ordinarily fits that description. It is not true, however, that precedent on costs calls for the judgment requirement the Court ironly adopts today for attorney’s fees. Indeed, the first decision cited in the concurring opinion, Mansfield, C. & L. M. B. Co. v. Swan, 111 U. S. 379 (1884), see ante, at 611, tugs against the restrictive rule today’s decision installs.

In Mansfield, plaintiffs commenced a contract action in state court. Over plaintiffs’ objections, defendants successfully removed the suit to federal court. Plaintiffs prevailed on the merits there, and defendants obtained review here. See 111 U. S., at 380-381. This Court determined, on its own motion, that federal subject-matter jurisdiction was absent from the start. Based on that determination, the Court reversed the lower court’s judgment for plaintiffs. Worse than entering and leaving this Courthouse equally “empty-handed,” ante, at 614 (concurring opinion), the plaintiffs in Mansfield were stripped of the judgment they had won, including the “judicial finding... of the merits” in their favor, ante, at 613 (concurring opinion). The Mansfield plaintiffs did, however, achieve this small consolation: The Court awarded them costs here as well as below. Recognizing that defendants had “prevailed]” in a “formal and nominal sense,” the Mansfield Court nonetheless concluded that “[i]n a true and proper sense” defendants were “the losing and not the prevailing party.” Ill U. S., at 388.

While Mansfield easts doubt on the present majority’s “formal and nominal” approach, that decision does not con*631sider whether costs would be in order for the plaintiff who obtains substantial relief, but no final judgment. Nor does “a single cose” on which the concurring opinion today relies, ante, at 613 (emphasis in original).6 There are, however, enlightening analogies. In multiple instances, state high courts have regarded plaintiffs as prevailing, for costs taxation purposes, when defendants’ voluntary conduct, mooting the suit, provided the relief that plaintiffs sought.7 The con*632curring opinion labors unconvincingly to distinguish these state-law cases.8 A similar federal practice has been observed in cases governed by Federal Rule of Civil Procedure 64(d), the default rule allowing costs “to the prevailing party unless the court otherwise directs.” See 10 C. Wright, A. Miller, & M. Kane, Federal Practice and Procedure §2667, pp. 187-188 (2d ed. 1983) (When “the defendant alters its conduct so that plaintiff’s claim [for injunctive relief] becomes moot before judgment is reached, costs may be allowed [under Rule 54(d)] if the court finds that the changes *633were the result, at least in part, of plaintiff’s litigation.”) (citing, inter alia, Black Hills Alliance v. Regional Forester, 526 F. Supp. 257 (SD 1981)).

In short, there is substantial support, both old and new, federal and state, for a costs award, “in [the court’s] discretion,” supra, at 625, n. 1, to the plaintiff whose suit prompts the defendant to provide the relief plaintiff seeks.

C

Recognizing that no practice set in stone, statute, rule, or precedent, see infra, at 643, dictates the proper construction of modern civil rights fee-shifting prescriptions, I would “assume . . . that Congress intends the words in its enactments to carry ‘their ordinary, contemporary, common meaning.’ ” Pioneer, 507 U.S., at 388 (defining “excusable neglect”) (quoting Perrin v. United States, 444 U.S. 37, 42 (1979) (defining “bribery”)); see also, e. g., Sutton v. United Air Lines, Inc., 527 U.S. 471, 491 (1999) (defining “substantially” in light of ordinary usage); Rutledge v. United States, 517 U. S. 292, 299-300, n. 10 (1996) (similarly defining “in concert”). In everyday use, “prevail” means “gain victory by virtue of strength or superiority: win mastery: triumph.” Webster’s Third New International Dictionary 1797 (1976). There are undoubtedly situations in which an individual’s goal is to obtain approval of a judge, and in those situations, one cannot “prevail” short of a judge’s formal declaration. In a piano competition or a figure skating contest, for example, the person who prevails is the person declared winner by the judges. However, where the ultimate goal is not an arbiter’s approval, but a favorable alteration of actual circumstances, a formal declaration is not essential. Western democracies, for instance, “prevailed” in the Cold War even though the Soviet Union never formally surrendered. Among television viewers, John F. Kennedy “prevailed” in the first debate with Richard M. Nixon during the 1960 Presidential contest, even though moderator Howard K. Smith *634never declared a winner. See T. White, The Making of the President 1960, pp. 298-294 (1961).

A lawsuit’s ultimate purpose is to achieve actual relief from an opponent. Favorable judgment may be instrumental in gaining that relief Generally, however, “the judicial decree is not the end but the means. At the end of the rainbow lies not a judgment, but some action (or cessation of action) by the defendant . . . .” Hewitt v. Helms, 482 U. S. 755, 761 (1987). On this common understanding, if a party reaches the “sought-after destination,” then the party “prevails” regardless of the “route taken.” Hennigan v. Ouachita Parish School Bd., 749 F. 2d 1148, 1153 (CA5 1985).

Under a fair reading of the FHAA and ADA provisions in point, I would hold that a party “prevails” in “a true and proper sense,” Mansfield, 111 U. S., at 388, when she achieves, by instituting litigation, the practical relief sought in her complaint. The Court misreads Congress, as I see it, by insisting that, invariably, relief must be displayed in a judgment, and correspondingly that a defendant’s voluntary action never suffices. In this case, Buekhannon’s purpose in suing West Virginia officials was not narrowly to obtain a judge’s approbation. The plaintiffs’ objective was to stop enforcement of a rule requiring Buckhannon to evict residents like centenarian Dorsey Pierce as the price of remaining in business. If Buckhannon achieved that objective on account of the strength of its case, see supra, at 628 — if it succeeded in keeping its doors open while housing and caring for Ms. Pierce and others similarly situated — then Buckhan-non is properly judged a party who prevailed.

III

As the Courts of Appeals have long recognized, the catalyst rule suitably advances Congress’ endeavor to place private actions, in civil rights and other legislatively defined areas, securely within the federal law enforcement arsenal.

*635The catalyst rule stemmed from modern legislation extending civil rights protections and enforcement measures. The Civil Rights Act of 1964 included provisions for fee awards to “prevailing parties” in Title II (public accommodations), 42 U. S. C. §2000a-3(b), and Title YII (employment), §2000e-5(k), but not in Title VI (federal programs). The provisions’ central purpose was “to promote vigorous enforcement” of the laws by private plaintiffs; although using the two-way term “prevailing party,” Congress did not make fees available to plaintiffs and defendants on equal terms. Christiansburg Garment Co. v. EEOC, 434 U.S. 412, 417, 421 (1978) (under Title VII, prevailing plaintiff qualifies for fee award absent “special circumstances,” but prevailing defendant may obtain fee award only if plaintiff’s suit is “frivolous, unreasonable, or without foundation”).

Once the 1964 Act came into force, courts commenced to award fees regularly under the statutory authorizations, and sometimes without such authorization. See Alyeska Pipeline Service Co. v. Wilderness Society, 421 U.S. 240, 262, 270-271, n. 46 (1975). In Alyeska, this Court reaffirmed the “American Rule” that a court generally may not award attorney’s fees without a legislative instruction to do so. See id., at 269. To provide the authorization Alyeska required for fee awards under Title VI of the 1964 Civil Rights Act, as well as under Reconstruction Era civil rights legislation, 42 U. S. C. §§ 1981-1983, 1985, 1986 (1994 ed. and Supp. V), and certain other enactments, Congress passed the Civil Rights Attorney’s Fees Awards Act of 1976,42 U.S. C. § 1988 (1994 ed. and Supp. V).

As explained in the Reports supporting § 1988, civil rights statutes vindicate public policies “of the highest priority,” S. Rep. No. 94-1011, p. 3 (1976) (quoting Newman v. Piggie Park Enterprises, Inc., 390 U.S. 400, 402 (1968) (per curiam)), yet “depend heavily upon private enforcement,” S. Rep. No. 94-1011, at 2. Persons who bring meritorious civil rights claims, in this light, serve as “private attorneys *636general.” Id., at 5; H. R. Rep. No. 94-1558, p. 2 (1976). Such suitors, Congress recognized, often “cannot afford legal counsel.” Id., at 1. They therefore experience “severe hardship]” under the “American Rule.” Id., at 2. Congress enacted §1988 to ensure that nonaffluent plaintiffs would have “effective access” to the Nation’s courts to enforce civil rights laws. Id., at l.9 That objective accounts for the fee-shifting provisions before the Court in this case, prescriptions of the FHAA and the ADA modeled on § 1988. See supra, at 624-625, n. 1.

Under the catalyst rule that held sway until today, plaintiffs who obtained the relief they sought through suit on genuine claims ordinarily qualified as “prevailing parties,” so that courts had discretion to award them their costs and fees. Persons with limited resources were not impelled to “wage total law” in order to assure that their counsel fees would be paid. They could accept relief, in money or of another kind, voluntarily proffered by a defendant who sought to avoid a recorded decree. And they could rely on a judge then to determine, in her equitable discretion, whether counsel fees were warranted and, if so, in what amount.10

*637Congress appears to have envisioned that very prospect. The Senate Report on the 1976 Civil Rights Attorney’s Fees Awards Act states: “[F]or purposes of the award of counsel fees, parties may be considered to have prevailed when they vindicate rights through a consent judgment or without formally obtaining relief.” S. Rep. No. 94-1011, at 5 (emphasis added). In support, the Report cites cases in which parties recovered fees in the absence of any court-conferred relief.11 *638The House Report corroborates: “[A]fter a complaint is filed, a defendant might voluntarily cease the unlawful practice. A court should still award fees even though it might conclude, as a matter of equity, that no formal relief, such as an injunction, is needed.” H. R. Rep. No. 94-1558, at 7 (emphases added). These Reports, Courts of Appeals have observed, are hardly ambiguous. Compare ante, at 607-608 (“legislative history ... is at best ambiguous”), with, e. g., Dunn v. The Florida Bar, 889 F. 2d 1010,1013 (CA11 1989) (legislative history “evinces a clear Congressional intent” to permit award “even when no formal judicial relief is obtained” (internal quotation marks omitted)); Robinson v. Kimbrough, 652 F. 2d 458, 465 (CA5 1981) (same); American Constitutional Party v. Munro, 650 F. 2d 184, 187 (CA9 1981) (Senate Report “directs” fee award under catalyst rule). Congress, I am convinced, understood that “ ‘[v]ic-tory' in a civil rights suit is typically a practical, rather than a strictly legal matter.” Exeter-West Greenwich Regional School Dist. v. Pontarelli, 788 F. 2d 47, 51 (CA1 1986) (citation omitted).

IV

The Court identifies several “policy arguments” that might warrant rejection of the catalyst rule. See ante, at 608-610. A defendant might refrain from altering its conduct, fearing liability for fees as the price of voluntary action. See ante, at 608. Moreover, rejection of the catalyst rule has limited impact: Desisting from the challenged conduct will not render a case moot where damages are sought, and even when the plaintiff seeks only equitable relief, a defendant’s voluntary cessation of a challenged practice does not render the case moot “unless it is ‘absolutely clear that the allegedly wrongful behavior could not reasonably be expected to recur.’” Ante, at 609 (quoting Friends of Earth, Inc., 528 *639U. S., at 189). Because a mootness dismissal is not easily achieved, the defendant may be impelled to settle, negotiating fees less generous than a court might award. See ante, at 609. Finally, a catalyst rule would “require analysis of the defendant’s subjective motivations,” and thus protract the litigation. Ibid.

The Court declines to look beneath the surface of these arguments, placing its reliance, instead, on a meaning of “prevailing party” that other jurists would scarcely recognize as plain. See ante, at 608. Had the Court inspected the “policy arguments” listed in its opinion, I doubt it would have found them impressive.

In opposition to the argument that defendants will resist change in order to stave off an award of fees, one could urge that the catalyst rule may lead defendants promptly to comply with the law’s requirements: the longer the litigation, the larger the fees. Indeed, one who knows noncompliance will be expensive might be encouraged to conform his conduct to the legal requirements before litigation is threatened. Cf. Hylton, Fee Shifting and Incentives to Comply with the Law, 46 Vand. L. Rev. 1069, 1121 (1993) (“fee shifting in favor of prevailing plaintiffs enhances both incentives to comply with legal rules and incentives to settle disputes”). No doubt, a mootness dismissal is unlikely when recurrence of the controversy is under the defendant’s control. But, as earlier observed, see swpra, at 636, why should this Court’s fee-shifting rulings drive a plaintiff prepared to accept adequate relief, though out-of-court and unrecorded, to litigate on and on? And if the catalyst rule leads defendants to negotiate not only settlement terms but also allied counsel fees, is that not a consummation to applaud, not deplore?

As to the burden on the court, is it not the norm for the judge to whom the case has been assigned to resolve fee disputes (deciding whether an award is in order, and if it is, the amount due), thereby clearing the case from the calendar? If factfinding becomes neeessary under the catalyst *640rule, is it not the sort that “the district courts, in their fact-finding expertise, deal with on a regular basis”? Baumgartner v. Harrisburg Housing Auth., 21 F. 3d 541, 548 (CA3 1994). Might not one conclude overall, as Courts of Appeals have suggested, that the catalyst rule “saves judicial resources,” Paris v. Department of Housing and Urban Development, 988 F. 2d 236, 240 (CA1 1998), by encouraging “plaintiffs to discontinue litigation after receiving through the defendant’s acquiescence the remedy initially sought”? Morris v. West Palm Beach, 194 F. 3d 1203, 1207 (CA11 1999).

The concurring opinion adds another argument against the catalyst rule: That opinion sees the rule as accommodating the “extortionist” who obtains relief because of “greater strength in financial resources, or superiority in media manipulation, rather than superiority in legal merit” Ante, at 617, 618 (emphasis in original). This concern overlooks both the character of the rule and the judicial superintendence Congress ordered for all fee allowances. The catalyst rule was auxiliary to fee-shifting statutes whose primary purpose is “to promote the vigorous enforcement” of the civil rights laws. Christiansburg Garment Co., 434 U.S., at 422. To that end, courts deemed the conduct-altering catalyst that counted to be the substance of the case, not merely the plaintiff’s atypically superior financial resources, media ties, or political clout. See supra, at 628. And Congress assigned responsibility for awarding fees not to automatons unable to recognize extortionists, but to judges expected and instructed to exercise “discretion.” See supra, at 624-625, n. 1. So viewed, the catalyst rule provided no berth for nuisance suits, see Hooper, 37 F. 3d, at 292, or “thinly disguised forms of extortion,” Tyler v. Corner Constr. Corp., 167 F. 3d 1202, 1206 (CA8 1999) (citation omitted).12

*641V

As to our attorney’s fee precedents, the Court correctly observes, “[w]e have never had occasion to decide whether the term ‘prevailing party' allows an award of fees under the ‘catalyst, theory,”’ and “there is language in our cases supporting both petitioners and respondents.” Ante, at 603, n. 5. It bears emphasis, however, that in determining whether fee shifting is in order, the Court in the past has placed greatest weight not on any “judicial imprimatur ante, at 605, but on the practical impact of the lawsuit.13 In Maher v. Gagne, 448 U.S. 122 (1980), in which the Court held fees could be awarded on the basis of a consent decree, the opinion nowhere relied on the presence of a formal judgment. See supra, at 629; infra, at 642-643, n. 14. Some years *642later, in Hewitt v. Helms, 482 U.S. 755 (1987), the Court suggested that fees might be awarded the plaintiff who “obtained] relief without [the] benefit of a formal judgment.” Id., at 760. The Court explained: “If the defendant, under the pressure of the lawsuit, pays over a money claim before the judicial judgment is pronounced,” or “if the defendant, under pressure of [a suit for declaratory judgment], alters his conduct (or threatened conduct) towards the plaintiff,” i. e., conduct “that was the basis for the suit, the plaintiff will have prevailed.” Id., at 761. I agree, and would apply that analysis to this case.

The Court posits a “‘merit’ requirement of our prior cases.” Ante, at 606. Maher, however, affirmed an award of attorney’s fees based on a consent decree that “did not purport to adjudicate [plaintiff’s] statutory or constitutional claims.” 448 U. S., at 126, n. 8. The decree in Maher “explicitly stated that 'nothing [therein was] intended to constitute an admission of fault by either party.’” Ibid. The catalyst rule, in short, conflicts with none of “our prior holdings,” ante, at 605.14

*643* * *

The Court states that the term “prevailing party” in fee-shifting statutes has an “accepted meaning.” Ante, at 608. If that is so, the “accepted meaning” is not the one the Court today announces. It is, instead, the meaning accepted by every Court of Appeals to address the catalyst issue before our 1987 decision in Hewitt, see supra, at 626, n. 4, and disavowed since then only by the Fourth Circuit, see supra, at 627, n. 5. A plaintiff prevails, federal judges have overwhelmingly agreed, when a litigated judgment, consent decree, out-of-court settlement, or the defendant’s voluntary, posteomplaint payment or change in conduct in fact affords redress for the plaintiff’s substantial grievances.

When this Court rejects the considered judgment prevailing in the Circuits, respect for our colleagues demands a co*644gent explanation. Today’s decision does not provide one. The Court’s narrow construction of the words "prevailing party” is unsupported by precedent and unaided by history or logic. Congress prescribed fee-shifting provisions like those included in the FHAA and ADA to encourage private enforcement of laws designed to advance civil rights. Fidelity to that purpose calls for court-awarded fees when a private party’s lawsuit, whether or not its settlement is registered in court, vindicates rights Congress sought to secura I would so hold and therefore dissent from the judgment and opinion of the Court.

*719Justice Stevens

delivered the opinion of the Court.

The Civil Rights Attorney’s Fees Awards Act of 1976 (Fees Act) provides that “the court, in its discretion, may allow the prevailing party ... a reasonable attorney’s fee” in *720enumerated civil rights actions. 90 Stat. 2641, 42 U. S. C. § 1988. In Maher v. Gagne, 448 U. S. 122 (1980), we held that fees may be assessed against state officials after a case has been settled by the entry of a consent decree. In this case, we consider the question whether attorney’s fees must be assessed when the case has been settled by a consent decree granting prospective relief to the plaintiff class but providing that the defendants shall not pay any part of the prevailing party’s fees or costs. We hold that the District Court has the power, in its sound discretion, to refuse to award fees.

I

The petitioners are the Governor and other public officials of the State of Idaho responsible for the education and treatment of children who suffer from emotional and mental handicaps. Respondents are a class of such children who have been or will be placed in petitioners’ care.1

On August 4, 1980, respondents commenced this action by filing a complaint against petitioners in the United States District Court for the District of Idaho. The factual allegations in the complaint described deficiencies in both the educational programs and the health care services provided respondents. These deficiencies allegedly violated the United States Constitution, the Idaho Constitution, four *721federal statutes, and certain provisions of the Idaho Code. The complaint prayed for injunctive relief and for an award of costs and attorney’s fees, but it did not seek damages.

On the day the complaint was filed, the District Court entered two orders, one granting the respondents leave to proceed in forma pauperis, and a second appointing Charles Johnson as their next friend for the sole purpose of instituting and prosecuting the action. At that time Johnson was employed by the Idaho Legal Aid Society, Inc., a private, nonprofit corporation that provides free legal services to qualified low-income persons.2 Because the Idaho Legal Aid Society is prohibited from representing clients who are capable of paying their own fees,3 it made no agreement requiring any of the respondents to pay for the costs of litigation or the legal services it provided through Johnson. Moreover, the special character of both the class and its attorney-client relationship with Johnson explains why it did not enter into any agreement covering the various contingencies that might arise during the course of settlement negotiations of a class action of this kind.

Shortly after petitioners filed their answer, and before substantial work had been done on the case, the parties entered into settlement negotiations. They were able to reach agreement concerning that part of the complaint relating to educational services with relative ease and, on October 14, 1981, entered into a stipulation disposing of that part of the case. The stipulation provided that, each party would bear its “own attorney’s fees and costs thus far incurred.” App. *72254. The District Court promptly entered an order approving the partial settlement.

Negotiations concerning the treatment claims broke down, however, and the parties filed cross-motions for summary judgment. Although the District Court dismissed several of respondents’ claims, it held that the federal constitutional claims raised genuine issues of fact to be resolved at trial. Thereafter, the parties stipulated to the entry of a class certification order, engaged in discovery, and otherwise prepared to try the case in the spring of 1988.

In March 1983, one week before trial, petitioners presented respondents with a new settlement proposal. As respondents themselves characterize it, the proposal “offered virtually all of the injunctive relief [they] had sought in their complaint.” Brief for Respondents 5. See App. 89. The Court of Appeals agreed with this characterization, and further noted that the proposed relief was “more than the district court in earlier hearings had indicated it was willing to grant.” 748 F. 2d 648, 650 (CA9 1984). As was true of the earlier partial settlement, however, petitioners’ offer included a provision for a waiver by respondents of any claim to fees or costs.4 Originally, this waiver was unacceptable to the Idaho Legal Aid Society, which had instructed Johnson to reject any settlement offer conditioned upon a waiver of fees, but Johnson ultimately determined that his ethical obligation to his clients mandated acceptance of the proposal. The parties conditioned the waiver on approval by the District Court.5

*723After the stipulation was signed, Johnson filed a written motion requesting the District Court to approve the settlement “except for the provision on costs and attorney’s fees,” and to allow respondents to present a bill of costs and fees for consideration by the court. App. 87. At the oral argument on that motion, Johnson contended that petitioners’ offer had exploited his ethical duty to his clients — that he was “forced,” by an offer giving his clients “the best result [they] could have gotten in this court or any other court,” to waive his attorney’s fees.6 The District Court, however, evaluated the waiver in the context of the entire settlement and rejected the ethical underpinnings of Johnson’s argument. Explaining that although petitioners were “not willing to concede that they were obligated to [make the changes in their practices required by the stipulation], . . . they were willing to do them as long as their costs were outlined and they didn’t face additional costs,” it concluded that “it doesn’t violate any ethical considerations for an attorney to give up his attorney fees in the interest of getting a better bargain for his client[s].” Id., at 93. Accordingly, the District Court ap*724proved the settlement and denied the motion to submit a costs bill.

When respondents appealed from the order denying attorney’s fees and costs, petitioners filed a motion requesting the District Court to suspend or stay their obligation to comply with the substantive terms of the settlement. Because the District Court regarded the fee waiver as a material term of the complete settlement, it granted the motion.7 The Court of Appeals, however, granted two emergency motions for stays requiring enforcement of the substantive terms of the consent decree pending the appeal. More dramatically, after ordering preliminary relief, it invalidated the fee waiver and left standing the remainder of the settlement; it then instructed the District Court to “make its own determination of the fees that are reasonable” and remanded for that limited purpose. 743 F. 2d, at 652.

In explaining its holding, the Court of Appeals emphasized that Rule 23(e) of the Federal Rules of Civil Procedure gives the court the power to approve the terms of all settlements of class actions,8 and that the strong federal policy embodied in *725the Fees Act normally requires an award of fees to prevailing plaintiffs in civil rights actions, including those who have prevailed through settlement.9 The court added that “[w]hen attorney’s fees are negotiated as part of a class action settlement, a conflict frequently exists between the class lawyers’ interest in compensation and the class members’ interest in relief.” 743 F. 2d, at 651-652. “To avoid this conflict,” the Court of Appeals relied on Circuit precedent which had “disapproved simultaneous negotiation of settlements and attorney’s fees” absent a showing of “unusual circumstances.” Id., at 652.10 In this case, the Court of Appeals found no such “unusual circumstances” and therefore held that an agreement on fees “should not have been a part of the settlement of the claims of the class.” Ibid. It concluded:

“The historical background of both Rule 23 and section 1988, as well as our experience since their enactment, compel the conclusion that a stipulated waiver of all attorney’s fees obtained solely as a condition for obtaining relief for the class should not be accepted by the court.” Ibid.

*726The importance of the question decided by the Court of Appeals, together with the conflict between its decision and the decisions of other Courts of Appeals,11 led us to grant certio-rari. 471 U. S. 1098 (1985). We now reverse.

HH Í — i

The disagreement between the parties and amici as to what exactly is at issue in this case makes it appropriate to put certain aspects of the case to one side in order to state precisely the question that the case does present.

To begin with, the Court of Appeals’ decision rested on an erroneous view of the District Court’s power to approve settlements in class actions. Rule 23(e) wisely requires court approval of the terms of any settlement of a class action, but the power to approve or reject a settlement negotiated by the parties before trial does not authorize the court to require the parties to accept a settlement to which they have not agreed. Although changed circumstances may justify a court-ordered modification of a consent decree over the objections of a party after the decree has been entered,12 and the District Court *727might have advised petitioners and respondents that it would not approve their proposal unless one or more of its provisions was deleted or modified, Rule 23(e) does not give the court the power, in advance of trial, to modify a proposed consent decree and order its acceptance over either party’s objection.13 The options available to the District Court were essentially the same as those available to respondents: it could have accepted the proposed settlement; it could have rejected the proposal and postponed the trial to see if a different settlement could be achieved; or it could have decided to try the case. The District Court could not enforce the settlement on the merits and award attorney’s fees anymore than it could, in a situation in which the attorney had negotiated a large fee at the expense of the plaintiff class, preserve the fee award and order greater relief on the merits. The question we must decide, therefore, is whether the District Court had a duty to reject the proposed settlement because it included a waiver of statutorily authorized attorney’s fees.

That duty, whether it takes the form of a general prophylactic rule or arises out of the special circumstances of this case, derives ultimately from the Fees Act rather than from the strictures of professional ethics. Although respondents contend that Johnson, as counsel for the class, was faced with an “ethical dilemma” when petitioners offered him relief greater than that which he could reasonably have expected to obtain for his clients at trial (if only he would stipulate to a waiver of the statutory fee award), and although we recognize Johnson’s conflicting interests between pursuing relief for the class and a fee for the Idaho Legal Aid Society, we do *728not believe that the “dilemma” was an “ethical” one in the sense that Johnson had to choose between conflicting duties under the prevailing norms of professional conduct. Plainly, Johnson had no ethical obligation to seek a statutory fee award. His ethical duty was to serve his clients loyally and competently.14 Since the proposal to settle the merits was more favorable than the probable outcome of the trial, Johnson’s decision to recommend acceptance was consistent with the highest standards of our profession. The District Court, therefore, correctly concluded that approval of the settlement involved no breach of ethics in this case.

The defect, if any, in the negotiated fee waiver must be traced not to the rules of ethics but to the Fees Act.15 Fol*729lowing this tack, respondents argue that the statute must be construed to forbid a fee waiver that is the product of “coercion.” They submit that a “coercive waiver” results when the defendant in a civil rights action (1) offers a settlement on the merits of equal or greater value than that which plaintiffs could reasonably expect to achieve at trial but (2) conditions the offer on a waiver of plaintiffs’ statutory eligibility for attorney’s fees. Such an offer, they claim, exploits the ethical obligation of plaintiffs’ counsel to recommend settlement in order to avoid defendant’s statutory liability for its opponents’ fees and costs.16

The question this case presents, then, is whether the Fees Act requires a district court to disapprove a stipulation seeking to settle a civil rights class action under Rule 23 when the offered relief equals or exceeds the probable outcome at trial but is expressly conditioned on waiver of statutory eligibility for attorney’s fees. For reasons set out below, we are not persuaded that Congress has commanded that all such settlements must be rejected by the District Court. Moreover, on the facts of record in this case, we are satisfied that the Dis*730trict Court did not abuse its discretion by approving the fee waiver.

Ill

The text of the Fees Act provides no support for the proposition that Congress intended to ban all fee waivers offered in connection with substantial relief on the merits.17 On the contrary, the language of the Act, as well as its legislative history, indicates that Congress bestowed on the “prevailing party” (generally plaintiffs18) a statutory eligibility for a discretionary award of attorney’s fees in specified civil rights actions.19 It did not prevent the party from waiving this eli*731gibility anymore than it legislated against assignment of this right to an attorney, such as effectively occurred here. Instead, Congress enacted the fee-shifting provision as “an integral part of the remedies necessary to obtain” compliance with civil rights laws, S. Rep. No. 94-1011, p. 5 (1976), to further the same general purpose — promotion of respect for civil rights — that led it to provide damages and injunctive relief. The statute and its legislative history nowhere suggest that Congress intended to forbid all waivers of attorney’s fees — even those insisted upon by a civil rights plaintiff in exchange for some other relief to which he is indisputably not entitled20 — anymore than it intended to bar a concession on damages to secure broader injunctive relief. Thus, while it is undoubtedly true that Congress expected fee shifting to attract competent counsel to represent citizens deprived of their civil rights,21 it neither bestowed fee awards upon attor*732neys nor rendered them nonwaivable or nonnegotiable; instead, it added them to the arsenal of remedies available to combat violations of civil rights, a goal not invariably inconsistent with conditioning settlement on the merits on a waiver of statutory attorney’s fees.22

In fact, we believe that a general proscription against negotiated waiver of attorney’s fees in exchange for a settlement on the merits would itself impede vindication of civil rights, at least in some cases, by reducing the attractiveness of settlement. Of particular relevance in this regard is our recent decision in Marek v. Chesny, 473 U. S. 1 (1985). In that case, which admittedly was not a class action and therefore did not implicate the court’s approval power under Rule 23(e), we specifically considered and rejected the contention that civil rights actions should be treated differently from other civil actions for purposes of settlement. As The Chief Justice explained in his opinion for the Court, the settlement of litigation provides benefits for civil rights plain*733tiffs as well as defendants and is consistent with the purposes of the Fees Act:

“There is no evidence, however, that Congress, in considering § 1988, had any thought that civil rights claims were to be on any different footing from other civil claims insofar as settlement is concerned. Indeed, Congress made clear its concern that civil rights plaintiffs not be penalized for ‘helping to lessen docket congestion’ by settling their cases out of court. See H. R. Rep. No. 94-1558, supra, at 7.

“. . . Some plaintiffs will receive compensation in settlement where, on trial, they might not have recovered, or would have recovered less than what was offered. And, even for those who would prevail at trial, settlement will provide them with compensation at an earlier date without the burdens, stress, and time of litigation. In short, settlements rather than litigation will serve the interests of plaintiffs as well as defendants.” 473 U. S., at 10.

To promote both settlement and civil rights, we implicitly acknowledged in Marek v. Chesny the possibility of a tradeoff between merits relief and attorney’s fees when we upheld the defendant’s lump-sum offer to settle the entire civil rights action, including any liability for fees and costs.

In approving the package offer in Marek v. Chesny we recognized that a rule prohibiting the comprehensive negotiation of all outstanding issues in a pending case might well preclude the settlement of a substantial number of cases:

“If defendants are not allowed to make lump-sum offers that would, if accepted, represent their total liability, they would understandably be reluctant to make settlement offers. As the Court of Appeals observed, ‘many a defendant would be unwilling to make a binding settlement offer on terms that left it exposed to liability for attorney’s fees in whatever amount the court might *734fix on motion of the plaintiff.’ 720 F. 2d, at 477.” Id., at 6-7.

See White v. New Hampshire Dept. of Employment Security, 455 U. S. 445, 454, n. 15 (1982) (“In considering whether to enter a negotiated settlement, a defendant may have good reason to demand to know his total liability from both damages and fees”).

Most defendants are unlikely to settle unless the cost of the predicted judgment, discounted by its probability, plus the transaction costs of further litigation, are greater than the cost of the settlement package. If fee waivers cannot be negotiated, the settlement package, must either contain an attorney’s fee component of potentially large and typically uncertain magnitude, or else the parties must agree to have the fee fixed by the court. Although either of these alternatives may well be acceptable in many cases, there surely is a significant number in which neither alternative will be as satisfactory as a decision to try the entire case.23

The adverse impact of removing attorney’s fees and costs from bargaining might be tolerable if the uncertainty introduced into settlement negotiations were small. But it is not. The defendants’ potential liability for fees in this kind of litigation can be as significant as, and sometimes even more significant than, their potential liability on the merits. This proposition is most dramatically illustrated by the fee awards *735of district courts in actions seeking only monetary relief.24 Although it is more difficult to compare fee awards with the cost of injunctive relief, in part because the cost of such relief is seldom reported in written opinions, here too attorney’s fees awarded by district courts have “frequently outrun the economic benefits ultimately obtained by successful litigants.” 122 Cong. Rec. 31472 (1976) (remarks of Sen. Kennedy).25 Indeed, in this very case “[c]ounsel for defendants view[ed] the risk of an attorney’s fees award as the most significant liability in the case.” Brief for Defendants in Support of Approval of Compromise in Jeff D. v. Evans, No. 80-4091 (D. Idaho), p. 5. Undoubtedly there are many other civil rights actions in which potential liability for attorney’s fees may overshadow the potential cost of relief on the merits and darken prospects for settlement if fees cannot be negotiated.

The unpredictability of attorney’s fees may be just as important as their magnitude when a defendant is striving to fix its liability. Unlike a determination of costs, which ordinarily involve smaller outlays and are more susceptible of calculation, see Marek v. Chesny, 473 U. S., at 7, “[tjhere is no precise rule or formula” for determining attorney’s fees, *736Hensley v. Eckerhart, 461 U. S. 424, 436 (1983).26 Among other considerations, the district court must determine what hours were reasonably expended on what claims, whether that expenditure was reasonable in light of the success obtained, see id., at 436, 440, and what is an appropriate hourly rate for the services rendered. Some District Courts have also considered whether a “multiplier” or other adjustment is appropriate. The consequence of this succession of necessarily judgmental decisions for the ultimate fee award is inescapable: a defendant’s liability for his opponent’s attorney’s fees in a civil rights action cannot be fixed with a sufficient degree of confidence to make defendants indifferent to their exclusion from negotiation.27 It is therefore not implausible to anticipate that parties to a significant number of civil rights cases will refuse to settle if liability for attorney’s fees remains open,28 thereby forcing more cases to trial, unnec*737essarily burdening the judicial system, and disserving civil rights litigants. Respondents’ own waiver of attorney’s fees and costs to obtain settlement of their educational claims is eloquent testimony to the utility of fee waivers in vindicating civil rights claims.29 We conclude, therefore, that it is not *738necessary to construe the Fees Act as embodying a general rule prohibiting settlements conditioned on the waiver of fees in order to be faithful to the purposes of that Act.30

h — i <1

The question remains whether the District Court abused its discretion in this case by approving a settlement which included a complete fee waiver. As noted earlier, Rule 23(e) wisely requires court approval of the terms of any settlement *739of a class action. The potential conflict among members of the class — in this case, for example, the possible conflict between children primarily interested in better educational programs and those primarily interested in improved health care — fully justifies the requirement of court approval.

The Court of Appeals, respondents, and various amici supporting their position, however, suggest that the court’s authority to pass on settlements, typically invoked to ensure fair treatment of class members, must be exercised in accordance with the Fees Act to promote the availability of attorneys in civil rights cases. Specifically, respondents assert that the State of Idaho could not pass a valid statute precluding the payment of attorney’s fees in settlements of civil rights cases to which the Fees Act applies. See Brief for Respondents 24, n. 22. From this they reason that the Fees Act must equally preclude the adoption of a uniform statewide policy that serves the same end, and accordingly contend that a consistent practice of insisting on a fee waiver as a condition of settlement in civil rights litigation is in conflict with the federal statute authorizing fees for prevailing parties, including those who prevail by way of settlement.31 Remarkably, there seems little disagreement on these points. Petitioners and the amici who support them never suggest that the district court is obligated to place its stamp of approval on every settlement in which the plaintiffs’ attorneys have agreed to a fee waiver. The Solicitor General, for ex*740ample, has suggested that a fee waiver need not be approved when the defendant had “no realistic defense on the merits,” Brief for United States as Amicus Curiae Supporting Reversal 23, n. 9; see id., at 26-27,32 or if the waiver was part of a “vindictive effort ... to teach counsel that they had better not bring such cases,” Tr. of Oral Arg. 22.

We find it unnecessary to evaluate this argument, however, because the record in this case does not indicate that Idaho has adopted such a statute, policy, or practice. Nor does the record support the narrower proposition that petitioners’ request to waive fees was a vindictive effort to deter attorneys from representing plaintiffs in civil rights suits against Idaho. It is true that a fee waiver was requested and obtained as a part of the early settlement of the education claims, but we do not understand respondents to be challenging that waiver, see Tr. of Oral Arg. 31-32, and they have not offered to prove that petitioners’ tactics in this case merely implemented a routine state policy designed to frustrate the objectives of the Fees Act. Our own examination of the record reveals no such policy.

*741In light of the record, respondents must — to sustain the judgment in their favor — confront the District Court’s finding that the extensive structural relief they obtained constituted an adequate quid pro quo for their waiver of attorney’s fees.33 The Court of Appeals did not overturn this finding. Indeed, even that court did not suggest that the option of rejecting the entire settlement and requiring the parties either to try the case or to attempt to negotiate a different settlement would have served the interests of justice. Only by making the unsupported assumption that the respondent class was entitled to retain the favorable portions of the settlement while rejecting the fee waiver could the Court of Appeals conclude that the District Court had acted unwisely.

What the outcome of this settlement illustrates is that the Fees Act has given the victims of civil rights violations a powerful weapon that improves their ability to employ counsel, to obtain access to the courts, and thereafter to vindicate their rights by means of settlement or trial. For aught that appears, it was the “coercive” effect of respondents’ statutory right to seek a fee award that motivated petitioners’ exceptionally generous offer. Whether this weapon might be even more powerful if fee waivers were prohibited in cases like this is another question,34 but it is in any event a question *742that Congress is best equipped to answer. Thus far, the Legislature has not commanded that fees be paid whenever a case is settled. Unless it issues such a command, we shall rely primarily on the sound discretion of the district courts to appraise the reasonableness of particular class-action settlements on a case-by-case basis, in the light of all the relevant circumstances.35 In this case, the District Court did not *743abuse its discretion in upholding a fee waiver which secured broad injunctive relief, relief greater than that which plaintiffs could reasonably have expected to achieve at trial.36

The judgment of the Court of Appeals is reversed.

It is so ordered.

Justice Brennan,

with whom Justice Marshall and Justice Blackmun join, dissenting.

Ultimately, enforcement of the laws is what really counts. It was with this in mind that Congress enacted the Civil Rights Attorney’s Fees Awards Act of 1976, 42 U. S. C. § 1988 (Act or Fees Act). Congress authorized fee shifting to improve enforcement of civil rights legislation by making it easier for victims of civil rights violations to find lawyers willing to take their cases. Because today’s decision will make it more difficult for civil rights plaintiffs to obtain legal assistance, a result plainly contrary to Congress’ purpose, I dissent.

I

The Court begins its analysis by emphasizing that neither the language nor the legislative history of the Fees Act supports “the proposition that Congress intended to ban all fee waivers offered in connection with substantial relief on the merits.” Ante, at 730. I agree. There is no evidence that *744Congress gave the question of fee waivers any thought at all. However, the Court mistakenly assumes that this omission somehow supports the conclusion that fee waivers are permissible. On the contrary, that Congress did not specifically consider the issue of fee waivers tells us absolutely nothing about whether such waivers ought to be permitted. It is black letter law that “[i]n the absence of specific evidence of Congressional intent, it becomes necessary to resort to a broader consideration of the legislative policy behind th[e] provision . . . .” Brooklyn Savings Bank v. O’Neil, 324 U. S. 697, 706 (1945); see also 2A C. Sands, Sutherland on Statutory Construction §§54.01-54.03 (4th ed. 1984). We must interpret the statute in the way that is most consistent with Congress’ broader purpose; a result which is “plainly at variance with the policy of the legislation as a whole,” Ozawa v. United States, 260 U. S. 178, 194 (1922), cannot be correct. Watt v. Western Nuclear, Inc., 462 U. S. 36, 56 (1983) (statute should not be interpreted “to produce a result at odds with the purposes underlying the statute” but rather “in a way that will further Congress’ overriding objective”); 2A Sands, swpra, §46.07; see also United States v. Freeman, 3 How. 556, 565 (1845); Sorrells v. United States, 287 U. S. 435, 446 (1932); United States v. Brown, 333 U. S. 18, 25-26 (1948); Lynch v. Overholser, 369 U. S. 705, 710 (1962); Perry v. Commerce Loan Co., 383 U. S. 392, 399-400 (1966) (quoting United States v. American Trucking Assns., 310 U. S. 534, 543 (1940)); United States v. Campos-Serrano, 404 U. S. 293, 298 (1971). Accordingly, the first and most important question to be asked is what Congress’ purpose was in enacting the Fees Act. We must then determine whether conditional fee waivers are consistent with this pin-pose.

I — I I — I

The Court asserts that Congress authorized fee awards to further the same general purpose — promotion of respect for civil rights — that led it to provide damages and injunctive *745relief.” Ante, at 731. The attorney’s fee made available by the Act, we are told, is simply an addition to “the arsenal of remedies available to combat violations of civil rights.” Ante, at 732.

Obviously, the Fees Act is intended to “promote respect for civil rights.” Congress would hardly have authorized fee awards in civil rights cases to promote respect for the securities laws. But discourse at such a level of generality is deceptive. The question is how did Congress envision that awarding attorney’s fees would promote respect for civil rights? Without a clear understanding of the way in which Congress intended for the Fees Act to operate, we cannot even begin responsibly to go about the task of interpreting it. In theory, Congress might have awarded attorney’s fees as simply an additional form of make-whole relief, the threat of which would “promote respect for civil rights” by deterring potential civil rights violators. If this were the case, the Court’s equation of attorney’s fees with damages would not be wholly inaccurate. However, the legislative history of the Fees Act discloses that this is not the case. Rather, Congress provided fee awards to ensure that there would be lawyers available to plaintiffs who could not otherwise afford counsel, so that these plaintiffs could fulfill their role in the federal enforcement scheme as “private attorneys general,” vindicating the public interest.1

*746Before the late 1960’s, the concept of fee shifting in public interest litigation was virtually nonexistent. In Newman v. Piggie Park Enterprises, Inc., 390 U. S. 400 (1968) (per curiam), this Court was called upon to interpret the attorney’s fee provision of Title II of the then recently enacted Civil Rights Act of 1964, 42 U. S. C. § 2000a-3(b). We held that a prevailing plaintiff should ordinarily recover fees unless special circumstances rendered such an award unjust. Noting that “[w]hen the Civil Rights Act of 1964 was passed, it was evident that enforcement would prove difficult and that the Nation would have to rely in part upon private litigation as a means of securing broad compliance with the law,” we recognized that “[a] Title II suit is thus private in form only.” Newman, 390 U. S., at 401. If a plaintiff obtains relief, he “does so not for himself alone but also as a ‘private attorney general,’ vindicating a policy that Congress considered of the highest priority.” Id., at 402 (footnote omitted). We recognized further that the right to recover attorney’s fees was conferred by Congress to ensure that this private public-enforcement mechanism would operate effectively:

“If successful plaintiffs were routinely forced to bear their own attorneys’ fees, few aggrieved parties would be in a position to advance the public interest by invoking the injunctive powers of the federal courts. Congress therefore enacted the provision for counsel fees — not simply to penalize litigants who deliberately advance arguments they know to be untenable but, more broadly, to encourage individuals injured by racial discrimination *747to seek judicial relief under Title II.” Ibid, (footnote omitted).

Newman interpreted the fee provision of Title II as intended to bridge the gap between the desire of an individual who has been deprived of a federal right to see that right vindicated and the financial ability of that individual to do so. More importantly, Newman recognized that Congress did not erect this bridge solely, or even primarily, to confer a benefit on such aggrieved individuals. Rather, Congress sought to capitalize on the happy coincidence that encouraging private actions would, in the long run, provide effective public enforcement of Title II. By ensuring that lawyers would be willing to take Title II cases, Congress made the threat of a lawsuit for violating Title II real, thereby deterring potential violators.

After Newman, lower courts — invoking their equitable powers to award attorney’s fees — adopted a similar rationale to award fees in cases brought under civil rights statutes that did not contain express provisions for attorney’s fees. See, e. g., Stolberg v. Members of Board of Trustees for State Colleges of Conn., 474 F. 2d 485 (CA2 1973) (42 U. S. C. § 1983); Donahue v. Staunton, 471 F. 2d 475 (CA7 1972), cert. denied, 410 U. S. 955 (1973) (same); Lee v. Southern Home Sites Corp., 444 F. 2d 143 (CA5 1971) (42 U. S. C. §1982). See generally Derfner, One Giant Step: The Civil Rights Attorney’s Fees Awards Act of 1976, 21 St. Louis U. L. J. 441, 443, and nn. 9-22 (1977) (citing cases). In May 1975, this Court in Alyeska Pipeline Service Co. v. Wilderness Society, 421 U. S. 240, ruled that the equitable powers of the federal courts did not authorize fee awards on the ground that a case served the public interest. Although recognizing that “Congress has opted to rely heavily on private enforcement to implement public policy and to allow counsel fees so as to encourage private litigation,” the Court held that “congressional utilization of the private-attorney-general concept can in no sense be construed as a grant of authority to the *748Judiciary ... to award attorneys’ fees whenever the courts deem the public policy furthered by a particular statute important enough to warrant the award.” Id., at 263. Instead, the Court ruled, only Congress could authorize awarding fees as a means of encouraging private actions in the name of public policy. Id., at 269-271.

In the wake of Alyeska, Congress acted to correct “anomalous gaps” in the availability of attorney’s fees to enforce civil rights laws, S. Rep. No. 94-1011, p. 1 (1976) (hereafter S. Rep.).2 See H. R. Rep. No. 94-1558, p. 2 (1976) (hereafter H. R. Rep.); 122 Cong. Rec. 31472 (1976) (remarks of Sen. Kennedy). Testimony at hearings on the proposed legislation disclosed that civil rights plaintiffs, “a vast majority of [whom] cannot afford legal counsel,” H. R. Rep. 1, were suffering “very severe hardships because of the Alyeska decision,” id., at 2. The unavailability of fee shifting made it impossible for legal aid services, “already short of resources,” to bring many lawsuits, and, without much possibility of compensation, private attorneys were refusing to take civil rights cases. Id., at 3. See generally Hearings on the Effect of Legal Fees on the Adequacy of Representation before the Subcommittee on Representation of Citizen Interests of the Senate Committee on the Judiciary, 93d Cong., 1st Sess., pts. 1-4 (1973). Congress found that Alyeska had a “devastating” impact on civil rights litigation, and it concluded that the need for corrective legislation was “compelling.” H. R. Rep. 3; see also, 122 Cong. Rec., supra, at 31471 (remarks of Sen. Scott), 31472 (remarks of Sen. Kennedy).

Accepting this Court’s invitation, see Alyeska, swpra, at 269-271, Congress passed the Fees Act in order to reestablish the Newman regime under which attorney’s fees were awarded as a means of securing enforcement of civil rights laws by ensuring that lawyers would be willing to *749take civil rights cases. The legislative history manifests this purpose with monotonous clarity. For instance, the Report of the House Judiciary Committee notes “The effective enforcement of Federal civil rights statutes depends largely on the efforts of private citizens. Although some agencies of the United States have civil rights responsibilities, their authority and resources are limited.” H. R. Rep. 1. The Report explains, quoting from Newman, that a plaintiff who obtains relief in a private lawsuit “‘does so not for himself alone but also as a “private attorney general,” vindicating a policy that Congress considered of the highest’ importance.” Id., at 2 (quoting 390 U. S., at 402). The Report then describes the intended scope and operation of the Fees Act, before concluding:

“The application of these standards will insure that reasonable fees are awarded to attract competent counsel in cases involving civil and constitutional rights, while avoiding windfalls to attorneys. The effect of [the Fees Act] will be to promote the enforcement of the Federal civil rights acts, as Congress intended, and to achieve uniformity in those statutes and justice for all citizens.” H. R. Rep. 9.

These same themes are prominent in the Senate Report:

“The purpose and effect of [the Fees Act] are simple— it is designed to allow courts to provide the familiar remedy of reasonable counsel fees to prevailing parties in suits to enforce the civil rights acts which Congress has passed since 1866. ... All of these civil rights laws depend heavily upon private enforcement, and fee awards have proved an essential remedy if private citizens are to have a meaningful opportunity to vindicate the important Congressional policies which these laws contain.” S. Rep. 2.

The Senate Report quotes the same language from Newman as the House Report, explaining that “fees are an integral *750part of the remedy necessary to achieve compliance with our statutory policies.” Id., at 3. After citing existing fee-shifting provisions, the Report sets out the Committee’s finding that “[t]hese fee shifting provisions have been successful in enabling vigorous enforcement of modern civil rights legislation, while at the same time limiting the growth of the enforcement bureaucracy.” Id., at 4. The Report then concludes: “If our civil rights laws are not to become mere hollow pronouncements which the average citizen cannot enforce, we must maintain the traditionally effective remedy of fee shifting in these cases.” Id., at 6.

The floor debates, which were extensive, also are replete with similar expressions; I set out but a few examples. Senator Tunney, who sponsored the original version of the Pees Act, stated to the Senate:

“The problem of unequal access to the courts in order to vindicate congressional policies and enforce the law is not simply a problem for lawyers and courts. Encouraging adequate representation is essential if the laws of this Nation are to be enforced. Congress passes a great deal of lofty legislation promising equal rights to all.

“Although some of these laws can be enforced by the Justice Department or other Federal agencies, most of the responsibility for enforcement has to rest upon private citizens, who must go to court to prove a violation of the law. . . . But without the availability of counsel fees, these rights exist only on paper. Private citizens must be given not only the rights to go to court, but also the legal resources. If the citizen does not have the resources, his day in court is denied him; the congressional policy which he seeks to vindicate goes unvindicated; and the entire Nation, not just the individual citizen, suffers.” 122 Cong. Rec. 33313 (1976).

Senator Kennedy, who sponsored the amended version of the Fees Act that was actually passed, made the same point somewhat more succinctly:

*751“Long experience has demonstrated . . . that Government enforcement alone cannot accomplish [compliance with the civil rights laws]. Private enforcement of these laws by those most directly affected must continue to receive full congressional support. Fee shifting provides a mechanism which can give full effect to our civil rights laws, at no added cost to the Government.” Id., at 31472.

But perhaps it was Representative Anderson, responding to a question from an opponent of the Fees Act, who summed up the reason for the legislation most effectively. He said:

“We are talking here about major civil rights laws. We have an obligation, it seems to me, as the representatives of the people, to make sure that those laws are enforced and we discharge that obligation when we make available a reasonable award of attorneys’ fees at the discretion of the court. Those of us who are interested in making sure that those laws are enforced . . . are simply abetting and aiding that process of law enforcement when we agree to the provisions of this bill.” Id., at 35116.

See also, e. g., id., at 31471 (remarks of Sen. Scott) (“Congress should encourage citizens to go to court in private suits to vindicate its policies and protect their rights”), 35128 (remarks of Rep. Seiberling).

h-1 I — I

As this review of the legislative history makes clear, then, by awarding attorney’s fees Congress sought to attract competent counsel to represent victims of civil rights violations.3 Congress’ primary purpose was to enable “private attorneys *752general” to protect the public interest by creating economic incentives for lawyers to represent them. The Court’s assertion that the Fees Act was intended to do nothing more than give individual victims of civil rights violations another remedy is thus at odds with the whole thrust of the legislation. Congress determined that the public as a whole has an interest in the vindication of the rights conferred by the civil rights statutes over and above the value of a civil rights remedy to a particular plaintiff.4

I have gone to great lengths to show how the Court mis-characterizes the purpose of the Fees Act because the Court’s error leads it to ask the wrong question. Having concluded that the Fees Act merely creates another remedy to vindicate the rights of individual plaintiffs, the Court asks whether negotiated waivers of statutory attorney’s fees are “invariably inconsistent” with the availability of such fees as a remedy for individual plaintiffs. Ante, at 732. Not surprisingly, the Court has little difficulty knocking down this frail straw man.5 But the proper question is whether permitting nego*753tiated fee waivers is consistent with Congress’ goal of attracting competent counsel. It is therefore necessary to consider the effect on this goal of allowing individual plaintiffs to negotiate fee waivers.

A

Permitting plaintiffs to negotiate fee waivers in exchange for relief on the merits actually raises two related but distinct questions. First, is it permissible under the Fees Act to negotiate a settlement of attorney’s fees simultaneously with the merits? Second, can the “reasonable attorney’s fee” guaranteed in the Act be waived? As a matter of logic, either of these practices may be permitted without also permitting the other. For instance, one could require bifurcated settlement negotiations of merits and fees but allow plaintiffs to waive their fee claims during that phase of the negotiations. Alternatively, one could permit simultaneous negotiation of fees and merits but prohibit the plaintiff from waiving statutory fees. This latter possibility exists because there is a range of “reasonable attorney’s fees” consistent with the Fees Act in any given case. Cf. Blum v. Stenson, 465 U. S. 886 (1984); Hensley v. Eckerhart, 461 U. S. 424, 433-437 (1983); H. R. Rep. 8-9; S. Rep. 6; see generally Johnson v. Georgia Highway Express, Inc., 488 F. 2d 714, 716-720 (CA5 1974) (listing relevant factors).6

More importantly, since simultaneous negotiation and waiver may have different effects on the congressional policy of encouraging counsel to accept civil rights cases, each practice must be analyzed independently to determine whether or *754not it is consistent with the Fees Act. Unfortunately, the Court overlooks the logical independence of simultaneous negotiation and waiver and assumes that there cannot be one without the other. See ante, at 734-738, and n. 28. As a result, the Court’s discussion conflates the different effects of these practices, and its opinion is of little use in coming to a fair resolution of this case. An independent examination leads me to conclude: (1) that plaintiffs should not be permitted to waive the “reasonable fee” provided by the Fees Act; but (2) that parties may undertake to negotiate their fee claims simultaneously with the merits so long as whatever fee the parties agree to is found by the court to be a “reasonable” one under the Fees Act.

B

1

It seems obvious that allowing defendants in civil rights cases to condition settlement of the merits on a waiver of statutory attorney’s fees will diminish lawyers’ expectations of receiving fees and decrease the willingness of lawyers to accept civil rights cases. Even the Court acknowledges “the possibility that decisions by individual clients to bargain away fee awards may, in the aggregate and in the long run, diminish lawyers’ expectations of statutory fees in civil rights cases.” Ante, at 741-742, n. 34. The Court tells us, however, that “[cjomment on this issue” is “premature at this juncture” because there is not yet supporting “documentation.” Ibid. The Court then goes on anyway to observe that “as a practical matter the likelihood of this circumstance arising is remote.” Ibid.

I must say that I find the Court’s assertions somewhat difficult to understand. To be sure, the impact of conditional fee waivers on the availability of attorneys will be less severe than was the restriction on fee awards created in Alyeska. However, that experience surely provides an indication of the immediate hardship suffered by civil rights claimants *755whenever there is a reduction in the availability of attorney’s fee awards.7 Moreover, numerous courts and commentators have recognized that permitting fee waivers creates disincentives for lawyers to take civil rights cases and thus makes it more difficult for civil rights plaintiffs to obtain legal assistance. See, e. g., Moore v. National Assn. of Securities Dealers, Inc., 246 U. S. App. D. C. 114, 133-134, 762 F. 2d 1093, 1112-1113 (Wald, J., concurring in judgment) id., at 138, 762 F. 2d, at 1117 (Wright, J., dissenting) (1985); Shadis v. Beal, 685 F. 2d 824, 830-831 (CA3), cert. denied sub nom. O’Bannon v. Shadis, 459 U. S. 970 (1982); Kraus, 29 Vill. L. Rev., at 625, 633-638; Comment, Settlement Offers Conditioned Upon Waiver of Attorneys’ Fees: Policy, Legal, and Ethical Considerations, 131 U. Pa. L. Rev. 793, 814-816 (1983); Committee on Professional and Judicial Ethics of the New York City Bar Association, Op. No. 80-94, reprinted in 36 Record of N. Y. C. B. A. 507, 508-509 (1981).

But it does not require a sociological study to see that permitting fee waivers will make it more difficult for civil rights plaintiffs to obtain legal assistance. It requires only common sense. Assume that a civil rights defendant makes a settlement offer that includes a demand for waiver of statutory attorney’s fees. The decision whether to accept or reject the *756offer is the plaintiff’s alone, and the lawyer must abide by the plaintiff’s decision. See, e. g., ABA, Model Rules of Professional Conduct 1.2(a) (1984); ABA, Model Code of Professional Responsibility EC 7-7 to EC 7-9 (1982).8 As a formal matter, of course, the statutory fee belongs to the plaintiff, ante, at 730, and n. 19, and thus technically the decision to waive entails a sacrifice only by the plaintiff. As a practical matter, however, waiver affects only the lawyer. Because “a vast majority of the victims of civil rights violations” have no resources to pay attorney’s fees, H. R. Rep. I,9 lawyers cannot hope to recover fees from the plaintiff and must depend entirely on the Fees Act for compensation.10 The plain*757tiff thus has no real stake in the statutory fee and is unaffected by its waiver. See Lipscomb v. Wise, 643 F. 2d 319, 320 (CA5 1981) (per curiam). Consequently, plaintiffs will readily agree to waive fees if this will help them to obtain other relief they desire.11 As summed up by the Legal Ethics Committee of the District of Columbia Bar:

“Defense counsel. . . are in a uniquely favorable position when they condition settlement on the waiver of the statutory fee: They make a demand for a benefit that the plaintiff’s lawyer cannot resist as a matter of ethics and one in which the plaintiff has no interest and therefore will not resist.” Op. No. 147, reprinted in 113 Daily Washington Reporter, supra n. 8, at 394.

Of course, from the lawyer’s standpoint, things could scarcely have turned out worse. He or she invested consid*758erable time and effort in the case, won, and has exactly nothing to show for it. Is the Court really serious in suggesting that it takes a study to prove that this lawyer will be reluctant when, the following week, another civil rights plaintiff enters his office and asks for representation? Does it truly require that somebody conduct a test to see that legal aid services, having invested scarce resources on a case, will feel the pinch when they do not recover a statutory fee?

And, of course, once fee waivers are permitted, defendants will seek them as a matter of course, since this is a logical way to minimize liability. Indeed, defense counsel would be remiss not to demand that the plaintiff waive statutory attorney’s fees. A lawyer who proposes to have his client pay more than is necessary to end litigation has failed to fulfill his fundamental duty zealously to represent the best interests of his client. Because waiver of fees does not affect the plaintiff, a settlement offer is not made less attractive to the plaintiff if it includes a demand that statutory fees be waived. Thus, in the future, we must expect settlement offers routinely to contain demands for waivers of statutory fees.12

The cumulative effect this practice will have on the civil rights bar is evident. It does not denigrate the high ideals that motivate many civil rights practitioners to recognize that lawyers are in the business of practicing law, and that, like other business people, they are and must be concerned with earning a living.13 The conclusion that permitting fee *759waivers will seriously impair the ability of civil rights plaintiffs to obtain legal assistance is embarrassingly obvious.

Because making it more difficult for civil rights plaintiffs to obtain legal assistance is precisely the opposite of what Congress sought to achieve by enacting the Fees Act, fee waivers should be prohibited. We have on numerous prior occasions held that “a statutory right conferred on a private party, but affecting the public interest, may not be waived or released if such waiver or release contravenes the statutory policy.” Brooklyn Savings Bank v. O’Neil, 324 U. S., at 704 (holding right to liquidated damages under Fair Labor Standards Act nonwaivable). See also, e. g., Boyd v. Grand Trank Western R. Co., 338 U. S. 263, 266 (1949) (holding venue provision of Federal Employers’ Liability Act non-waivable); Wilko v. Swan, 346 U. S. 427, 434-438 (1953) (holding void an agreement to arbitrate in lieu of judicial remedy provided by Securities Exchange Act); cf. James v. Home Construction Co. of Mobile, Inc., 689 F. 2d 1357, 1359 (CA11 1982) (implying a right of action for attorneys to seek fees under Truth-in-Lending Act to further congressional policies). This is simply straightforward application of the well-established principle that an agreement which is contrary to public policy is void and unenforceable. See Restatement (Second) of Contracts § 178 (1981); see also, Brooklyn Savings Bank v. O’Neil, supra, at 710; Grites, Inc. v. Prudential Insurance Co., 322 U. S. 408, 418 (1944); Weil v. Neary, 278 U. S. 160, 171-174 (1929); Woodstock Iron Co. v. Richmond & Danville Extension Co., 129 U. S. 643, 662-663 (1889).14

*7602

This all seems so obvious that it is puzzling that the Court reaches a different result. The Court’s rationale is that, unless fee waivers are permitted, “parties to a significant number of civil rights cases will refuse to settle . . . Ante, at 736. This is a wholly inadequate justification for the Court’s result.

First, the effect of prohibiting fee waivers on settlement offers is just not an important concern in the context of the Fees Act. I agree with the Court that encouraging settlements is desirable policy. But it is judicially created policy, applicable to litigation of any kind and having no special force in the context of civil rights cases.15 The congressional policy underlying the Fees Act is, as I have argued throughout, to create incentives for lawyers to devote time to civil rights cases by making it economically feasible for them to do so. Supra, at 745-753.16 As explained above, permitting fee *761waivers significantly undercuts this policy. Thus, even if prohibiting fee waivers does discourage some settlements, a judicial policy favoring settlement cannot possibly take precedence over this express congressional policy. We must implement Congress’ agenda, not our own.

In an attempt to justify its decision to elevate settlement concerns, the Court argues that settlement “provides benefits for civil rights plaintiffs as well as defendants and is consistent with the purposes of the Fees Act” because “ ‘[sjome plaintiffs will receive compensation in settlement where, on trial, they might not have recovered, or would have recovered less than what was offered.’” Ante, at 732-733 (quoting Marek v. Chesny, 473 U. S. 1, 10 (1985)); see also ante, at 731 (legislative history does not show that Congress intended to bar “even [waivers] insisted upon by a civil rights plaintiff in exchange for some other relief to which he is indisputably not entitled . . .”) (footnote omitted).

As previously noted, by framing the purpose of the Fees Act in very general terms, the Court merely obscures the proper focus of discussion. The Fees Act was designed to help civil rights plaintiffs in a particular way — by ensuring that there will be lawyers willing to represent them. The fact that fee waivers may produce some settlement offers that are beneficial to a few individual plaintiffs is hardly “consistent with the purposes of the Fees Act,” ante, at 733, if permitting fee waivers fundamentally undermines what Congress sought to achieve. Each individual plaintiff who waives his right to statutory fees in order to obtain additional relief for himself makes it that much more difficult for the next victim of a civil rights violation to find a lawyer willing or able to bring his case. As obtaining legal assistance becomes more difficult, the “benefit” the Court so magnani*762mously preserves for civil rights plaintiffs becomes available to fewer and fewer individuals, exactly the opposite result from that intended by Congress.

Moreover, I find particularly unpersuasive the Court’s apparent belief that Congress enacted the Fees Act to help plaintiffs coerce relief to which they are “indisputably not entitled.” See ante, at 731, 732. It may be that, in particular cases, some defendants’ fears of incurring liability for plaintiff’s attorney’s fees will give plaintiffs leverage to coerce relief they do not deserve. If so, this is an unfortunate cost of a statute intended to ensure that plaintiffs can obtain the relief to which they are entitled. And it certainly is not a result we must preserve at the expense of the central purpose of the Fees Act.

Second, even assuming that settlement practices are relevant, the Court greatly exaggerates the effect that prohibiting fee waivers will have on defendants’ willingness to make settlement offers. This is largely due to the Court’s failure to distinguish the fee waiver issue from the issue of simultaneous negotiation of fees and merits claims. Swpra, at 754. The Court’s discussion mixes concerns over a defendant’s reluctance to settle because total liability remains uncertain with reluctance to settle because the cost of settling is too high. See ante, at 734-737. However, it is a prohibition on simultaneous negotiation, not a prohibition on fee waivers, that makes it difficult for the defendant to ascertain his total liability at the time he agrees to settle the merits. Thus, while prohibiting fee waivers may deter settlement offers simply because requiring the defendant to pay a “reasonable attorney’s fee” increases the total cost of settlment, this is a separate issue altogether, and the Court’s numerous arguments about why defendants will not settle unless they can determine their total liability at the time of settlement, ante, at 734, 735, 736, are simply beside the point.17 With respect *763to a prohibition on fee waivers (and again merely assuming that effects on settlement are relevant), the sole question to be asked is whether the increased cost of settlement packages will prevent enough settlement offers to be a dispositive factor in this case.

The Court asserts, without factual support,18 that requiring defendants to pay statutory fee awards will prevent a “significant number” of settlements. Ante, at 734-735. It is, of course, ironic that the same absence of “documentation” which makes comment on the effects of permitting fee waivers “premature at this juncture,” ante, at 742, n. 34, does not similarly affect the Court’s willingness to speculate about what to expect if fee waivers are prohibited. Be that as it may, I believe that the Court overstates the extent to which prohibiting fee waivers will deter defendants from making settlement offers. Because the parties can negotiate a fee (or a range of fees) that is not unduly high and condition their settlement on the court’s approval of this fee, the magnitude *764of a defendant’s liability for fees in the settlement context need be neither uncertain nor particularly great.19 Against this, the defendant must weigh the risk of a nonnegotiated fee to be fixed by the court after a trial; as the Court reminds us, fee awards in this context may be very uncertain and, potentially, of very great magnitude. See ante, at 734-735, nn. 23, 24. Thus, powerful incentives remain for defendants to seek settlement. Moreover, the Court’s decision last Term in Marek v. Chesny, 473 U. S. 1 (1985), provides an additional incentive for defendants to make settlement offers, namely, the opportunity to limit liability for attorney’s fees if the plaintiff refuses the offer and proceeds to trial.

All of which is not to deny that prohibiting fee waivers will deter some settlements; any increase in the costs of settling will have this effect. However, by exaggerating the size and the importance of fee awards, and by ignoring the options available to the parties in settlement negotiations, the Court makes predictions that are inflated. An actual disincentive to settling exists only where three things are true: (1) the defendant feels he is likely to win if he goes to trial, in which case the plaintiff will recover no fees; (2) the plaintiff will agree to relief on the merits that is less costly to the defendant than litigating the case; and (3) adding the cost of a negotiated attorney’s fee makes it less costly for the defendant to litigate. I believe that this describes a very small class of cases — although, like the Court, I cannot “document” the assertion.

C

I would, on the other hand, permit simultaneous negotiation of fees and merits claims, since this would not contra*765vene the purposes of the Fees Act. Congress determined that awarding prevailing parties a “reasonable” fee would create necessary — and sufficient — incentives for attorneys to work on civil rights cases. Prohibiting plaintiffs from waiving statutory fees ensures that lawyers will receive this “reasonable” statutory fee. Thus, if fee waivers are prohibited, permitting simultaneous fees and merits negotiations will not interfere with the Act; the lawyer will still be entitled to and will still receive a reasonable attorney’s fee. Indeed, permitting simultaneous negotiations in such circumstances may even enhance the effectiveness of the Fees Act by making it easier for a lawyer to dispose of his cases more quickly. This frees up the lawyer’s time to take other cases and may enhance his reputation as an effective advocate who quickly obtains relief for clients.

IV

Although today’s decision will undoubtedly impair the effectiveness of the private enforcement scheme Congress established for civil rights legislation, I do not believe that it will bring about the total disappearance of “private attorneys general.” It is to be hoped that Congress will repair this Court’s mistake. In the meantime, other avenues of relief are available. The Court’s decision in no way limits the power of state and local bar associations to regulate the ethical conduct of lawyers. Indeed, several Bar Associations have already declared it unethical for defense counsel to seek fee waivers. See Committee on Professional Ethics of the Association of the Bar of the City of New York, Op. No. 82-80 (1985); District of Columbia Legal Ethics Committee, Op. No. 147, supra n. 8, 113 Daily Washington Law Reporter, at 389. Such efforts are to be commended and, it is to be hoped, will be followed by other state and local organizations concerned with respecting the intent of Congress and with protecting civil rights.

*766In addition, it may be that civil rights attorneys can obtain agreements from their clients not to waive attorney’s fees.20 Such agreements simply replicate the private market for legal services (in which attorneys are not ordinarily required to contribute to their client’s recovery21), and thus will enable civil rights practitioners to make it economically feasible — as Congress hoped — to expend time and effort litigating civil rights claims.

During the floor debates over passage of the Fees Act, Senator Hugh Scott reminded the Congress in terms that might well have been addressed to the Court today that “we must bear in mind at all times that rights that cannot be enforced through the legal process are valueless; such a situation breeds cynicism about the basic fairness of our judicial system. [We] must be vigilant to insure that our legal rights are not hollow ones.” 122 Cong. Rec. 31471 (1976).